Uptime: A Cloud Provider Sim
by RubyRack Games · 17 Jul, 2026
315 downloads
About the Game
You don't manage a datacenter from a spreadsheet. You walk it.
Uptime puts you on the floor of your own cloud provider, first person, hands on the hardware. You buy the servers and carry them to the rack. You seat them with a click. You run the cable off the spool yourself, find the port, watch it lay across the floor, and power the thing on. Then you sign your first customer and find out whether any of it holds.
Underneath the floor is a real simulation. Every server, switch, cable and port is a live entity in a deterministic engine, and the network genuinely switches and routes traffic across it. Power and cooling are budgets you can blow. Capacity is hardware that fills up. When something breaks, it breaks because the model says it should, and the fix is usually the one a real engineer would reach for. That was the whole goal: a build-it-with-your-hands tycoon sim that still holds up if you do this for a living.
Get it running, then keep it running. The moment traffic arrives, the simulation starts pushing back.
You're running the cloud, not a server room.
You own the full stack. The metal, the network, the services you sell, and the customers whose product lives on your kit. You're the thing they build on, so everything running on you is yours to keep alive, from the single box in the garage to a floor other companies depend on.
Your customers bet their business on you.
The tenants have names, and they hand you their product expecting it to stay up. Reputation is the currency underneath all of it. You earn it slowly and you can lose it in one bad night, and the bigger the customer, the harder that fall lands. Sign a deal you can't carry, pack one host too tight, sleep through an SLA breach at 2am, and the cost isn't only the money. It's the trust that was about to send you the next customer.
So every yes is a wager on whether you can carry the new weight without dropping what you already hold. Turning business away is sometimes the correct call. The tenants you keep won't sit quietly either. They surge, they haggle, they show up with deadlines. Keep them happy and they say so out loud, and they bring friends.
What you sell is infrastructure you actually run.
Customers don't rent "a server." They spin up the things a real cloud sells: virtual machines, object storage, managed databases, Kubernetes, load balancers, serverless functions. Each one is a service you operate. Behind every service is a control plane you have to keep alive and an agent on each host that does the work, and all of it competes for the same compute your customers are paying for.
That's where the network grows teeth. Cut a service off from its control plane and it starts to fail. Lean on one too hard and it buckles. Capacity runs out, the network can partition the very services you're selling, and the way you lay it all out is what separates a provider that holds from one that tips over.
The simulation is the game.
This is where the word "simulation" gets earned. Every device on your floor is a real entity in a deterministic engine, and every LED you see is reading its live state. A link saturates because the math saturated it. A rack goes dark because a fault domain you never thought about turned out to be a single point of failure.
Ports, cables, power draw, fault domains, oversubscription, link aggregation, spanning tree, BGP, mean-time-between-failures. These are the model, not labels stuck on a progress bar. Cable a redundant link the wrong way and spanning tree either saves you or, if you got it backwards, the whole floor drowns in a broadcast storm. You can ignore the depth and still win. You can also lean all the way into it.
The customers talk to you.
They live in a chat channel that reads like the ops Slack you've spent years in. They onboard, ask for more capacity, complain when latency spikes, and tell you in their own words when the change you made three minutes ago just took their service down.
Incidents don't ship with severity tags attached. Severity comes out of how many customers you're hurting and how badly. A flapping link nobody depends on is a shrug. That same link under your biggest tenant is a page in the middle of the night.
Two ways to read the same machine.
Never touched a rack? NPC mentors walk you in, and one key toggles jargon mode, flipping the whole interface between plain language and the real terminology over the exact same systems. "Spread these across separate failure zones" and "fix this single-AZ SPOF" are the same instruction. You pick up the real words by playing instead of reading a manual. Already know what oversubscription costs you? The depth was there the whole time. Turn the training wheels off.
Set the difficulty. The hard runs won't let you off.
One difficulty setting, Easy, Standard or Hard, re-tunes the entire simulation from a single neutral baseline, and you can move it mid-run. Push it up and the reputation ledger turns crueller: slower to earn, quicker to lose, less willing to forgive an incident. The authored hard scenarios set their own floor, so the slider can make them tougher but never softer.
Garage-to-Glory is one story, not the only one.
The campaign takes you from a single server in the garage to your own data center floor, growing through a basement build-out and into a room you fill and defend. That's one scenario. Run the hard-mode variant. Survive Data-Loss Day. Hold the line through Under Attack. Or drop into open sandbox and build with no rails at all. Scenarios are plain, readable files, so the community can write their own starts, disasters and constraints and drop them straight in.
This is Early Access. The data center floor is built to grow: add rows and cages to expand it, spread your customers across separate zones, and build real multi-region resilience by hand. More hardware, bigger scenarios and hyperscale scale are on the road ahead.
For the engineers.
You already know whether we mean it. Here's what's under the hood:
Typed ports and cables, with real LAG / link-aggregation across member links
Managed L2 and L3 switches forming an actual switching fabric, not one abstract "network" stat
RSTP that converges, elects a root bridge and blocks redundant ports; loop the network and it storms
Oversubscription that bites when you cheap out on uplinks
Fault domains and single points of failure that are computed, not painted on
Reachability chains from host to switch to demarc to uplink, gating every allocation
Every service split into a control plane and per-host data-plane agents; partition them and it goes headless
Power draw and wear modelling, so hardware ages and degrades instead of tripping a random death timer
Capacity modelled as workload shapes fitting host shapes rather than a scalar bar
Reputation as an asymmetric ledger, with a capacity-fit check before you ever sign
A deterministic engine: same seed, same run, every time, headless-testable and replay-stable
Built honestly.
A deterministic Rust engine runs the simulation, and the renderer only ever shows you what the engine already decided. Nothing gets faked for the camera. Scenarios and mods are human-readable files. Windows, macOS and Linux from day one.
Uptime is in Early Access. The full loop is here and playable now: campaign, sandbox, hard modes, and the customer and service simulation underneath them. Early Access is where the breadth grows on top of that, more places to build, more hardware, more scenarios, and steady polish across art, audio and onboarding, with the direction shaped by what players push on, the infra crowd most of all.
Features
First-person, hands-on - walk the floor, carry the hardware, run the cable. You're in the room, not hovering above it.
Cabling as a hero mechanic - run every link by hand, with typed connectors, enforced reach and bend radius, and a live edge that carries traffic. Pull it and something goes dark. The way Factorio makes belts feel good.
Customers as commitments - reputation earned slowly and lost in an instant, every signup a bet against the capacity you can see.
A real cloud to run - VMs, storage, databases, Kubernetes, load balancers and functions, each with a control plane and data-plane agents on your hosts.
A simulation that pushes back - ports, power, fault domains, spanning tree, oversubscription and BGP are the actual model.
Emergent incidents - severity comes from real customer impact, not scripted labels.
Readable for anyone, honest for the pros - one-key jargon toggle over the same systems, no manual required.
Difficulty you control - live Easy/Standard/Hard that re-tunes everything, and hard scenarios that won't soften.
Breadth of play - campaign, scenario variants, open sandbox, and drop-in community scenarios.
Deterministic and mod-friendly - same seed, same run, plain-text scenario files.
Screenshots
10 imagesVersion Information
Steam Patch Notes
Official update history
Added
- The Guide teaches the game. A top-left pane replaces the teaching arcs, quest panel and objective block, and each lesson says what a thing is, why it matters and how to do it.
- The Guide shows each lesson in full. It explains the idea and how to do the current step, and G switches it to one line and back, keeping your choice between sessions.
- Finishing a lesson opens the next one at once. A Done line above it names what now works, and the new lesson shows its steps straight away.
- The Guide shows inside the laptop. The foot of the laptop's page list shows the lesson, the current step and how to do it, and a refused action's reason.
- Help grows the longer you are stuck. After a minute stuck, a one-line Guide adds how to do the step, and Guided opens it in full at three and goes step by step at seven.
- New games ask "New to running servers?" Pick Guided, Light or Off, or change it later in the Guide section of Settings, where only the host can change it in multiplayer.
- Turning the Guide on mid-game skips what you have done. Switching from Off, or from Light to Guided, marks the lessons you have already finished as known.
- Guided garage starts hold new prospects until you are online. Garage to Glory, ARM Race, Hard mode, the demo and a sandbox garage send no new leads until you finish Getting online.
- Lessons open when they matter. Later lessons such as Under attack, Repairing a part, Moving up and Two of everything open with the event that needs them, never on a timer.
- Seven concept lessons explain the hard parts. They cover why a server says no, sharing processors, network cards, bonding cables, private networks, room in each zone and where the firewall goes.
- Urgent lessons interrupt. The attack lesson holds the pane while the attack lasts, and the lesson you were on comes back when it ends.
- Refusals and incidents point back at the lesson. In plain mode a refused action's reason shows in the Guide, one press from its lesson, and pops up instead while a device console covers the Guide.
- Skip or replay any lesson. The Lessons table in Knowledge replaces the Teaching Arcs list, and you can skip any lesson except the garage basics or replay one you have finished.
- Each scenario teaches its own lessons. A scenario that starts past the garage skips the garage lessons, and the Money lesson names the revenue and customers your next site asks for.
- Multiplayer shares one Guide. Lesson progress belongs to the world, every player sees the same lesson, and each player's marker follows their own carry or cable run.
Changed
- Technical mode still tells you what to do. The Guide's steps and the world markers give the same instructions as plain mode, naming where to go and what to press, not a status note.
- The Hints switch hides world markers only. With Hints off the Guide still explains every step in its pane, and the Teaching arcs switch is gone.
- Mira tells the story only. The Guide does the teaching now, so her teaching lines and ARM Race's setup quest ladder are gone.
- Mira's garage milestone comes at fifteen customers. It replaces her $20k cash line, which played as the game began, and her basement tease now follows a minute later.
- Concept cards no longer repeat a lesson. While the Guide runs, the rack, attack and SLA cards do not show when a lesson teaches the same idea.
- Incidents and messages name things by their names. Incident posts, diagnoses, bond labels and replies name zones, racks, customers, private networks and regions, so "AZ 2 outage" reads "AZ-2 outage".
- Anything unnamed reads one way. A server you never renamed reads srv-03 or server 3 until racked, then its pattern name, and a zone, customer or robot with no name reads like robot #3.
- Internet boxes are called internet boxes. In plain mode a new internet box reads internet box 3, not patch panel 3, and older saves correct the name when they load.
- New security boxes number by their id. A new firewall or scrubber can read scrub-03 where earlier boxes counted from 1, and the boxes you already own keep their names.
- The network map names zones and racks. Zone and rack clusters show their names, such as AZ-1, and the internet line names its gateway the way the rack does.
- A rename receipt names the device as it now reads. Under a pattern, that is the full name the device shows, not just the text you typed.
Fixed
- Device names no longer repeat location words. A server named az2-r1-web01 in zone AZ-2, rack AZ2-R1, reads az2-r1-u03-web01 instead of saying az2 three times, and nothing needs renaming.
- The server console's uplink list speaks plain mode. It names the switch each card plugs into in plain words when plain mode is on, instead of always showing names like sw-01.
- Name patterns show on every device surface. Rack plates, console titles, page rows, the Network view, the migration and segment pickers and the wallboard's busiest server now show the name your messages use.
- A rack or zone failure no longer freezes the game. Servers on the failed gear restart elsewhere all at once, and a large zone recovers without a long pause.
- Moving many workloads at once no longer stutters a big world. Drains, rebalances and growth that land hundreds of workloads finish in the same time with far less work per tick.
- Racking a server no longer stalls a big world. The game sends only what changed instead of the whole world, and after a load it no longer re-sends every customer as well.
- A power cut right after a save no longer empties it. The game writes each save fully to disk before it replaces the old one, and a backup never overwrites an older backup.
- A refused growth ask stays refused in a long game. Refusals no longer expire after about a thousand settled asks, so a customer you turned down waits until you install a server.
- Router port lights keep their traffic after an update. Gateway ports no longer blank for a moment when the router's settings change, and a multiplayer guest that lags behind still sees the change.
- Racking a router or internet box by hand respects rack power. The rack refuses gear past its breaker or onto a taken slot, the same as buying it straight into the rack.
- Mark all read clears only your own badges in multiplayer. A guest who marks everything read leaves the host's count alone, and every player sees their own unread count, on a dedicated server too.
- Loading a save no longer repeats old warnings. A save that is already ready to move up, or already short of power, loads quietly, and the move-up message appears once, not twice.
- The first scripted outage comes on time. In the garage and the demo the loose cable comes a few minutes after your first customer signs, and it hits a cable carrying a customer's work.
Notes
- Save compatibility. The save version is 64. v0.1.84 opens saves from earlier versions, but a v0.1.83 build cannot open a v0.1.84 save.
- Version lock. The game and its server now speak protocol 80, so a v0.1.83 player cannot join a v0.1.84 world, or the reverse.
- Load once to heal: lessons. On a save from before the Guide, lessons you already met and finished count as known, and a world past the garage skips the garage basics. Nothing moves.
- Load once to heal: device names. Internet boxes named patch panel N become internet box N, and a default name stops counting as one you typed. Names you typed stay as they are.
- Load once to heal: settings and robots. A save that had the teaching arcs switched off opens with the Guide Off, and a robot you named keeps its name.
- Mod API minor 8 is breaking. Hosts, switches, gateways, appliances and EDDs lose `display_name` and `display_name_plain`, and hosts lose `label`, so a minor 7 mod reading them is refused and reads `name` / `name_plain`.
- Mods: more names moved. `device_naming` loses `suffixed`, `kind_words`, `rack_rows`, `racks[].n` and the pool `word`, `host_ids` and `numbers`, and `stranded_vms[].host_display_name` becomes `host_name`. Rust mods use uptime-mod 0.9.0.
- Mods: story times are real seconds. Scripted-event `at_secs` and the knock-loose `delay_secs` become `at_real_secs` and `delay_real_secs`, real seconds at 1x, so a minor 7 scenario pack that sets them is refused.
- Mods: additive fields. Scenarios may set `lessons` and `hold_prospects_until_lesson`, a beat may set `replaces`, and the snapshot gains `name_plain` on named rows, `name` on patch panels and `stored_name` for rename prefills.
- API: the Guide. `GuidanceView.guide` (`GuideView`, `LessonView` with `args`, `LessonStepView`, `LessonNodeView`, `LessonEntryView`, `GuideAnnouncementView`, `GuidePointerView`) and `Hello.guide_mode` are new.
- API: Guide commands. `AcknowledgeLesson`, `SkipLesson`, `ReplayLesson` and the host-only `SetGuideMode` are new and game-client only, and `SetArcsEnabled` is removed.
- API: Guide error codes. Refusals can read `lesson_unknown`, `lesson_core`, `lesson_finished`, `lesson_not_finished`, `lesson_not_shown` or `guide_mode_unknown`.
- API: arcs removed. `GuidanceView` is now `hints_enabled` and `guide`, and its arc fields, `WorldSnapshot.arc_library`, `TeachingArcView`, `ArcConditionView` and `ClosedArcView` are gone.
- API: `hints_enabled` changed meaning. False now hides only the open lesson's world markers, in every mode, and the pane's own help stays.
- API: refusal recall. `CommandResult.recalled_lesson` names the lesson a refused command points its own player back at, and `GuideView.recalled` carries incident recalls only.
- API: names removed. `display_name`, `display_name_plain` and `HostView.label` are gone from host, switch, gateway, appliance, EDD and host console rows, along with `KindWordView`, `RackRowView`, `DeviceNamingView.suffixed` and `SuffixedNameView`.
- API: one name pair. Every named row carries `name` and `name_plain`, never empty, and `stored_name` on switch, gateway, appliance, EDD and host console rows is the rename prefill only.
- API: names renamed. `host_display_name` / `_plain` become `host_name` / `host_name_plain`, net-topo `display_name` / `_plain` become `name` / `name_plain`, and rack inspect occupant `label` becomes `name` / `name_plain`.
- API: names added. `name_plain` reaches racks, customers, segments, zones, regions, region detail and haulers, `name` / `name_plain` reach patch panels and the patch panel and EDD occupant rows, and `HaulerView.name` is never empty.
- API: names changed meaning. Incident, receipt, bond, refusal and cluster labels use displayed names, an unnamed device reads its default such as `srv-03`, and `RackView.name` reads `rack #N` instead of empty.
- API: pattern names drop repeats. A pattern `name` or `name_plain` leaves out location words an earlier placeholder already gave, so a server `az2-r1-web01` in rack `az2-r1` reads `az2-r1-u03-web01`.
- Wire: lighter frames. An ordinary delta can carry an empty `sectionsMsgpack`, read as the default snapshot, and `DeviceFlowView` leaves out zero numbers and empty `ports` with unchanged meaning.
- API: per-player unread. `ChannelView` gains `unread_by_player`, one `ChannelUnreadView` row per guest seat that has acted on the channel, omitted when empty, and `unread_for_player` is the host seat's count.
- API: uplink switch names. `HostUplinkView` gains `switch_name_plain`, the plain voice, and `switch_name` is the engineering half of the same pair, with the same text as before.
- API: name preview. The new `Request::PreviewNamePattern` answers `Response::NamePreview` with the name a pattern would give and the pattern a cleared level follows.
- API: message args. Segment, zone and region replies gain `segment_name`, `az_name` and `region_name` (plus `_eng`), and `ok-device-renamed`'s `name` is the name the device now shows.
Added
- Locate a server or switch. Locate on the Hosts page or the switch console blinks its front light and a rear-panel marker blue for every player, and the rack panel tags it Locating.
- Name your fleet your way. A Name pattern field with a live preview renames every device at once from parts such as {zone}, {rack}, {u} and {server}.
- Name patterns cascade. Set one for the fleet, a zone, a rack from its panel, or a pool on the Isolation page, and the nearest one wins.
- Pools name their servers. A pool pattern can use {admit} for the pool's network and {auto_number} for a number each server keeps while it stays in the pool.
- Racks get their own pattern. Build rack names from {zone}, {row} and a padded {n}, such as az1-R01.
- The comms panel sorts and remembers it. It opens latest first, you can sort by unread, latest or name, and your pick stays picked next time.
- Comms filters show what needs you. Needs attention holds a request until you decide and an alert until its cause is fixed, and tags each waiting message. Other filters show Incidents, Customers, Team or Prospects.
- Messages are tinted by type. Incidents are red, growth requests purple and prospects teal in their own channel, and the channel list is wider.
- Press F5 to quick-save. It saves like a manual save. In multiplayer only the host can save.
- The cable spool shows speeds and bend radius. You see the cable's top speed, the rated speed of the port you started from and the one you aim at, and the minimum bend radius.
- AOC cables come in orange. Pick it on the colour wheel when you lay an AOC run.
Changed
- Racked gear is named by where it sits. A server reads like az1-rack1-u03-srv-12 by default, the same in plain and engineering mode, and the name follows it when you move it or rename its rack.
- Incidents and alerts name the device by where it sits. Incident posts, alerts, refusals and the event feed use the same located name as the rack panel, so you can walk to the fault.
- No two devices share a name. A naming change that would make two devices read alike is refused, and building never is: the later device shows its kind and number appended.
- New racks get their own names. Racks are numbered Rack 1, Rack 2 and so on instead of all sharing the model name, and older saves renumber racks that still share one.
- Fixing the same fault twice in a day earns reputation once. A cable re-laid on the same ports is the same fault, and a second cable between the same two switches is its own.
- Racks can't be ordered past the room's free floor slots. This applies to the cart and to a direct buy, and crates you haven't installed yet count as taken.
- The ARM Race basement opens at $10k. It now matches Garage to Glory.
- Prospects that need an architecture you don't own say so. They no longer read Comfortable, and passing on them costs no reputation.
- Scrolling over the rack info panel scrolls the panel. The camera only moves when the cursor is over the rack.
Fixed
- Object storage always moves by copy. Buckets that grow, lose a server, change pool or spread to a new rack copy over the network and keep serving from their old disks until the copy lands.
- A bucket copy with no network path waits. It tries every server holding the data, tells you once, and resumes when a link appears.
- A growing bucket keeps every copy. When only one server has room for a whole copy, the other copy stripes across smaller servers, or growth waits for disk instead of dropping a copy.
- A demarc with no live handoff says so. One with lines bought but no working handoff reads NO LIVE HANDOFF on the network view and the Internet page instead of Online.
- Selling a rack always credits its catalog price. Renamed and renumbered racks sell at the price of the model you bought.
- Host pool changes queue their moves. Moving servers between pools no longer relocates every workload at once.
- An internet handoff with no line explains itself. The tooltip, the console and the rack panel say there is no internet line on it, instead of a bare Down.
- A handoff with no line no longer shows traffic. Its traffic figure reads zero while it reads Down.
- Internet handoffs keep up with their line. A handoff now updates within a tick when its line fails, recovers or is cancelled, and saves carrying a stale state are corrected on load.
Notes
- Save compatibility. The save version is 63. v0.1.83 opens saves from earlier versions, but a v0.1.82 build cannot open a v0.1.83 save.
- Version lock. The game and its server now speak protocol 79, so a v0.1.82 player cannot join a v0.1.83 world, or the reverse.
- Load once to heal. Stale internet handoffs are recomputed, racks that read alike get distinct names, each rack learns its model, and waiting prospects get a post. Nothing moves.
- Mod API minor 6. `locating`, `PortView.no_line`, `RackView.sku`, the timeline `msg_key`, `device_naming`, `LinesNotConnected` and the scenario `Prospects` channel are additive, so minor 5 mods keep working. Rust mods use uptime-mod 0.7.0.
- Mods: names changed meaning. Mods now read rack names as Rack 1 or the rack pattern instead of the model, and device names in messages as located names such as az1-rack1-u03-srv-12.
- API: rack names changed meaning. `RackView.name` is the rack's resolved name (its own name, or Rack 1, Rack 2 under the rack pattern) instead of the model, and `RackView.sku` carries the model.
- API: comms fields. `ChannelMessageView` gains `category`, `needs_attention`, `msg_key` and `msg_args`, and `ChannelView` gains `attention`. The `Prospects` channel kind and the `Prospect` and `CustomerRequest` message references are new.
- API: comms history. A channel's recent messages now also carry older messages that still need attention, beyond the newest 50.
- API: name cascade. `SetNameTemplate` is a new host-only command (game client only) with a level and target. The new `device_naming` section carries the patterns per level plus `suffixed`, the suffix a client appends.
- API: device names changed meaning. Device names in message args (`*_name`, `*_name_eng`) and events such as `WorkloadMoveLanded` now read az1-rack1-u03-srv-12 instead of srv-12 (Rack 1 U3). `display_name` fields are unchanged.
- API: new tag values. The incident timeline `kind` can read `ReputationWithheld`, timeline entries can carry `msg_key` and `msg_args`, and the command `SetDeviceLocate` is game-client only.
- API: new port and demarc values. `PortView.no_line` flags a handoff with no internet line, and `EddView.status` can read `LinesNotConnected`.
- Mods: F5 is reserved. A mod can no longer default a key to F5, which is now quick-save.
Added
- Reserve a whole rack to a private network. Servers racked there later join it, a server you pull out is open again, and Release on the rack panel or segment card opens the rack.
- A refused cable run marks the bad corner. The anchor at the tight bend turns red on the run you keep, so you can fix that stretch without hunting for it.
- Tray and J-hook corners are checked as you route. That includes a sharp turn where two tray sections meet, so a bad bend shows up before you plug in.
- A server in maintenance mode lists what cannot move off. Its console names each stuck workload and the reason, and the list clears when you take the server out of maintenance.
Changed
- Zones count room at a typical customer's size. One very large customer no longer makes a service read 0 everywhere, and the line below the table explains every zone with no room and why.
- The VM anti-affinity switch is gone. VMs always spread across servers, so the switch changed nothing.
- Bond, link-mode and bundle refusals use your language. Plain mode explains each one and what to try instead.
- Guests and dedicated-server players see save and load refusals in their own language. The same goes for picking content packs, which only the host can do.
- Incidents name devices by name. Labels such as "server #12" now use the device's name, the same as messages and the console.
- A breakout trunk takes one cable type. Later legs follow the first leg's type, and a cable that cannot seat in a leg's port is refused with an explanation.
- The single-player clock keeps pace when the game draws slowly. Below 4 frames a second, for example with the window on another workspace, the day no longer slows down.
- GPU 8x 7U and Mainframe 7U servers have 16 drive bays of 1 TB each. Their 16 TB used to split into drive sizes that do not exist.
- MRR says how it is counted. Each MRR figure says it is last hour's billing x 720, and the finance table shows burst vCPU billing as its own row.
- The Overselling CPU article explains burst billing. Packing servers tighter leaves less spare CPU for bursts, and bursts bill as extra revenue.
- The Off the Floor achievement is hidden for now. It could not unlock for cables routed through cable managers, and it returns with the cable neatness rework.
Fixed
- Multiplayer guests and slow clients no longer miss updates. When several world updates arrive at once after a stutter, lights, plugs and panels now show every change instead of only the last.
- A quiet world stops redrawing every cable. A cable change the game had already drawn no longer triggers a full cable pass again.
- WAN meters count only traffic that was carried. Traffic from powered-off or unreachable servers no longer counts as delivered, so inbound load, saturation warnings and attack impact match what moved.
- CDN storage growth can no longer overfill a server. Approving growth while one of the customer's sites is moving now books disk only where it fits.
- Loading a save no longer changes your world. It stops re-running checks and resetting incident times, the finance trend keeps its last hour, and usage not yet billed survives the save.
- An attack in progress keeps its escalation across save and load. Reloading no longer resets how long a flood has been leaking through.
- Network gear that fails again during a repair stays down. The first repair no longer brings it back early while a later fault is still being fixed.
- Clean failovers no longer count as harm. Customers that restart cleanly elsewhere after a server loses its power supplies are no longer marked as harmed.
- Two workload moves at once each get their notification. Moves for one customer in the same moment no longer hide one another.
- Network cards fitted in a server's box show in the event feed. A fitted card that cannot seat is refunded instead of staying on the pallet.
- The codex lists the switch pair split incident. It sits with the other incident types.
- Customers no longer feel lag from your own migrations. Background copies moving a workload stop counting as queueing on the links they share, so tenants stop complaining with no link full.
- The wall rack's bottom opening takes cables. Routing a cable out through the floor of a wall rack is no longer refused as too tight a bend.
- The placement ghost starts on the top free slot. It stays there until you move the mouse, then follows the cursor.
- Map groups open to show their servers. On a big fleet, click a zone or rack group to list its servers, and Back steps out. The Map opens framed on your servers.
- Walking into campus props no longer stalls the game. Trees, shelving, pallets, planters and floor vents use simpler collision, so they stop spiking the CPU or catching you on floor lips.
- Scenario prices and part-hour billing survive a save. A scenario's own price levels stay after a reload, and storage, load balancer and CDN revenue no longer comes up short after saving partway through an hour.
- Bonding onto an MLAG pair works from every console. Two MLAG pairs cabled back to back form one bundle, and routers, firewalls and appliances can bond onto a pair.
- A server keeps its fast bond when you add leads. It sends on the fastest bond or lead that reaches the network, keeps the rest on standby, and the console shows that rate.
- Object storage moves by copy. Shards on a server in maintenance, or spreading into a new rack, copy over the network instead of jumping to another server at once.
- A bucket stuck behind maintenance says why. When no other server has room for its data, the bucket says it is waiting and how much free disk the move needs.
- Dedicating servers no longer stacks a cluster on one server. Two workers of one cluster are planned onto different servers when you dedicate hosts.
Notes
- Save compatibility. The save version is 62. v0.1.82 opens saves from earlier versions, but a v0.1.81 build cannot open a v0.1.82 save.
- Version lock. The game and its server now speak protocol 78, so a v0.1.81 player cannot join a v0.1.82 world, or the reverse.
- Load once to heal. Very old saves drop work orders the game never built, and site fixtures such as the hauler, row widths and prices are filled in. Nothing moves.
- 7U servers on load. GPU 8x 7U and Mainframe 7U servers drop their surplus healthy drives on load. Storage stays the same and a failed drive is never removed.
- API: breakout refusals. A leg whose cable type differs from its trunk is refused with `error-breakout-leg-kind-mismatch`, and an explicit pick that cannot seat is refused with `breakout_kind_unusable`.
- API: trunk connector. A breakout trunk's port reports the connector of its first-laid leg's cable type, not whichever leg happens to be listed first.
- Mod API minor 5. `ReserveRackToSegment`, `ReleaseRackReservation` and `RackView.reserved_segment_id` are additive, so minor 4 mods keep working. Rust mods read them with uptime-mod 0.6.0.
- API: zone room changed meaning. `CapacityView.az_service_room` `room_units` and `unit_*` now count in the median in-use contract of each service, not the largest.
- API: anti-affinity ignored. `VmSettings.anti_affinity_enabled` and `SetVmSettings.anti_affinity` are still accepted, but they no longer change placement.
- API: refusal codes. Save, load and content-pack refusals to a guest or on a dedicated world carry code `host_only` and a `msg_key`.
- API: LAG refusal codes. `lag_max_members` and `lag_out_of_band_cable` now carry an empty code; read `msg_key` (`error-lag-max-members`, `error-lag-out-of-band`) instead.
- API: burst revenue. `revenue_breakdown.burst_last_hour_cents` is last hour's VM and managed database burst vCPU billing in cents. It is already inside VM revenue and MRR, so do not add it again.
- API: probe reason. A service's `reason_tag` can now read `WaitingOnDrain`, with `reason_key` `label-bucket-waits-on-drain` and a `size` in GB.
- API: cart kind. The cart view's `kind` can now read `Module`.
- Game client only. `HostConsoleSnapshot.drain_stuck` and the `CheckRouteBends` / `RouteBends` request are on the game snapshot and client, not the HTTP API.
Added
- Approve all and Refuse all on the Requests page. Approve all grants only growth asks and surges that fit. Poach threats, discount and compliance asks wait for you, and a toast gives the counts.
- The Capacity page shows room per zone. A Zones section lists how many more of each service fit in each zone for a new customer, and names the zone that is full or reserved.
- Plain mode explains refused actions. A wrong cable for a port, a full rack or a tight bend says why and what to try, and points to the new "Ports and cable types" codex article.
- Notifications have their own settings. You can turn off pop-ups, or only confirmations and routine updates. Refusals of your own actions, bulk reviews that left requests pending, failed saves and connection notices always show.
- Select all and Clear on the Isolation lists. Select all ticks only the hosts or customers your search and filters show, ready to assign or pin.
- Object storage buckets grow in bursts. Each bucket grows in uneven jumps on its own schedule, so some months take more disk and some less, while the long-run total still follows its contract.
- A bucket your fleet can't grow keeps serving and says so. Its row warns that it is out of storage with the figures, and storage requests count every disk you own, not one server.
Changed
Consoles and readings
- The host console shows real memory and storage. Resources rows read used and total, a failed memory module lowers the total, and percentages match the rack inspector, even past 100%.
- Hosts with a failed part read Degraded. The Hosts, Home, Capacity and Isolation pages stop calling them Healthy, and the Hosts page's fleet counts include them.
- A host in maintenance shows a blue light. The rack summary marks it apart from healthy and failed hosts.
- The shop lists each server's IOPS. You can compare disk speed before you buy.
- Host pools show storage. Isolation pool rows list storage used and total beside CPU and memory.
Customers and requests
- The customer page shows where and how big. Each host row shows its rack U, and every service shows its contracted storage, so object storage tenants no longer read blank.
- Auto-approved growth requests arrive already read. Approving one yourself earns reputation when you have the room and the customer is not fully down, even with auto-approve switched on.
- Customers never ask for more than your site could hold. Requests that need more servers still arrive, but none asks beyond what your building could fit.
- The auto-approve switch is also on the Requests page. It is the same company-wide setting as on the Customers page.
- The relocation breakdown adds up. It lists your cash on hand and any signup money that leaving onboarding customers return, so the rows sum to the capital you land with.
- Failing customers with no latency reading raise their own alert. The message says their requests are failing, and it opens the same incident a latency spike does.
Notifications and text
- Bursts of news fold into one card. Your own confirmations replace each other, and a run of the same event, such as cables laid, shows one counted card. In multiplayer, only your own lines fold.
- Cut-off names show in full on hover. Long device names and table cells end in an ellipsis instead of being clipped, and hovering shows the whole text.
- Read-only mods join saves you already have. Dashboards and alert mods can join an existing game. Mods that change your world need a new one, and one that gains commands later stays out.
Fixed
Large saves
- Large saves simulate their network about 45% faster each tick. Results are identical, so a busy campus keeps up with real time instead of falling behind.
- A failed server, switch or pulled cable freezes the game far less. The game recalculates routes once instead of twice, and faster, so the pause is about a third as long.
- Standing still in a big datacenter no longer stutters every few seconds. Racks and cables stop rebuilding when nothing changed, and walking past many cables no longer freezes a frame.
- Laptop screens, the delivery trolley and rack lights stop redoing work every tick. They update only when something they show changes, which removes a constant small stutter.
- Loading a save no longer shrinks a customer's contract. A busy fleet or a failed part made the load think the contract could never fit.
Gameplay
- A dark load balancer or CDN stops billing when it goes down. Its last slice before the outage still bills, and ports read down one tick sooner.
- Customers stop complaining about lag after an outage. A quiet customer no longer stays flagged, and Object Storage, CDN and load balancer customers report live latency, even with every host down.
- A refused cable run stays in your hands. When the game refuses a lay for a tight bend or its length, every clip stays put, even across racks, unless a menu or console is open.
- Approving bigger machines reserves the room at once. Two requests can no longer claim the same memory and CPU, a resized customer leaving frees only their own, and your power bill rises with the load.
- Storage and CDN growth requests reserve their disk when you approve them. A second request that no longer fits waits instead of being granted for the same space.
- Requests from departed customers leave the page. A customer who leaves takes their open growth asks and surges with them.
- Commands refused for too many requests say so. The game shows the refusal and undoes what it predicted.
- The latency chart shows a gap when a customer has no current reading. The line no longer drops to 0 ms.
- Mod warnings on load now show. A mod that was updated, reconfigured, held off or dropped from a save now says so when the save loads.
Notes
- Save compatibility. The save version stays 61, so v0.1.80 and v0.1.81 open each other's saves.
- Version lock. The game and its server now speak protocol 77. A v0.1.80 player cannot join a v0.1.81 world, or the reverse.
- Load once to heal. Stuck lag flags and requests from departed customers clear on load, and CDN sites under contract book their missing disk where it fits. Nothing moves and no customer leaves.
- New commands and event. `ApproveAllCustomerRequests` and `RefuseAllCustomerRequests` take no fields and are not available to mods. `CustomerRequestsBulkReviewed` reports the counts and the acting player.
- HTTP API fields added. `ServiceDemandView.storage_gb` and `RelocationOfferView.cash_dollars`.
- Game snapshot fields added. `HostSkuView.ssd_iops`, the host console's `mem_used_mb`, `mem_total_mb`, `ssd_used_gb` and `ssd_total_gb`, and the rack inspector's host `maintenance`. They are not on the HTTP API.
- Pool storage fields. `ReservationView` gains `ssd_used_gb` and `ssd_total_gb`, the storage in use and the effective storage summed over the pool's hosts, in the game and mod snapshots.
- API: room per zone. `CapacityView.az_service_room` has one row per zone per service in use, with `room_units`, unit size and `dedicated_hosts`. Key rows by zone and service, never by position.
- API: relocation capital. `RelocationWindowView` gains `clawback_dollars` and `new_capital_dollars`, the capital that lands. `projected_new_capital_dollars` reads before any clawback.
- API field that changed meaning. `CustomerView` `p99_latency_ms`, `p99_recent_ms` and `p50_latency_ms` are null when latency is stale, not only when absent. Treat null as no current latency, never zero.
- API: growth verdict changed meaning. `GrowthRequestView.fit_can_serve_now` is true for an ask needing no new room and false for one its contract cannot take. `pending_growth_requests` omits asks whose customer or service is gone.
- Mod API minor 4. The new snapshot fields, the bulk review event and `CustomerFailureSpike` are additive, so script mods built for minor 3 keep working. Rust mods read them with uptime-mod 0.5.0.
- Spike events. `CustomerFailureSpike` opens a spike with no current latency, and `CustomerLatencySpike.p99_ms` always holds a live reading.
- Mod approvals and warnings. A mod approving growth grants it and earns no reputation. The `script_not_applied` reason is retired, and `gameplay_not_applied` also covers a read-only mod that gained commands.
- Refusal message args. Refusal replies may carry `refusal_plain_key` and `refusal_article`. The bend refusal is coded `bend_radius_too_tight`, and its `anchor_index` counts the clips you placed and is absent on an engine-added corner.
- Command replies. Every reply to a notify command except a successful result now raises `command_failed`.
- More API fields. `CustomerRequestView.bulk_in_scope` says whether the bulk verbs act on a request. `CableLaid` and `CableRemoved` gain `actor`, and `fit_recommendation` no longer reads "No VM load" on an ask that serves now.
- API: storage reason changed meaning. `ServiceDemandView.reason` and `reason_key` can read Healthy with `ObjectStoreCapacityShort` and `label-bucket-held` for a serving bucket held for lack of disk. Then `PlacementFailureView.available` is the whole fleet's figure.
- First load of an older save. Every bucket moves onto its new growth path once during the load, which can make that one load slower.
- Host console percentages. `HostConsoleSnapshot` gains `mem_pct_effective` and `ssd_pct_effective`, uncapped, on the game snapshot only.
Changed
- Messages say where your gear is. Toasts, incidents, the event feed and #ops name a racked device with its rack and U, like "server 3 (Rack 30 U3)". A workload move shows one toast.
- You can order servers with their network cards fitted. Turn on "Fit network cards" when adding a server to the cart and pick a speed. The price includes the cards, and the box shows them.
- A network card is now a part you can move. You still buy cards from the server console in one click. You can also pull a card out and fit it into another server.
- Network cards wear out and fail. A dead card's ports go dark, its cables stay put, and an incident names the server until you replace the card.
- Selling a server with network cards in it pays for the cards too. Each card sells at its own resale value, racked or still boxed, and moving to a new site does the same.
- A server sends traffic over its fastest connected port. A fitted 10G or 25G card is used straight away instead of waiting in standby behind the onboard 1G port. Bonding is unchanged.
- A clean failover now earns reputation. When a server fails and nobody is harmed, the incident pays the full absorbed reward, about nine times the old amount. Sub-Five no longer fires when nobody was harmed.
- You can have 5 prospects waiting at once, up from 3. Existing games get the new limit when they load, and a sandbox keeps the limit you set.
- Active 100G and 400G DAC cables cost like other active DACs. They now run about 1.8x the passive price, and the cable picker offers them on 100G and 400G port pairs.
- Every extra power cord now adds heat. Each redundant cord adds 25 W of heat. A room just under its cooling limit with twin-fed gear can tip over after loading, so check your cooling headroom.
Fixed
- Power readings agree everywhere. A twin-fed internet box reads its full 41 W, unplugged gear and an unpowered gateway's optics draw nothing, and the rack inspector's devices add up to the rack total.
- Rack neatness reads the same everywhere. The rack inspector shows the score behind repair time and your neatness stats, counts racked routers and internet boxes at their slot, and judges power on nameplate draw.
- The near-breaker count matches the shop. The ops console counts a rack as near its breaker exactly when the shop would refuse more powered gear for it.
- A degraded link bundle reads right. Its event says how many links are up and how many it needs, instead of the garbled "3/1 members up".
- Pool and segment names are cleaned like every other name. Control characters you paste into them are stripped on create and rename.
- A switch failure always reaches your screen. A switch that failed on its own could keep showing as healthy until something else changed.
- Diagnosis confirms only the fault you checked. One failed server no longer confirms the others, a server and switch sharing a number stay apart, and an open incident's fault never reads cleared.
- The incident hint and diagnosis agree. When a cleared cause outscored a live one, the hint pointed at one cause and diagnosis rejected it. Both now lead with the same cause.
- An appliance that recovers after an inspection logs as info. It used to log as an error.
- SIEM logs keep every device's address distinct. Two devices could share one inside address in the security log, and each now has its own.
- A power outage holds every device in it. A device whose repair finishes during a breaker trip or zone outage stays down until the power returns, even when the outage runs past its expected end.
- A power cord no longer counts as a network cable. A server with one data cable no longer reads as having two, and an internet box's power cord no longer counts as cabling it.
- Evicting a new tenant shows the full fee. The fee includes the signup bonus and prepay you give back. If you cannot pay that back, the game refuses and shows what you owe.
- The Spread goal counts a switch pair as one. Two servers on a paired switch no longer complete it, matching how placement spreads workloads.
- Multiplayer guests must join before they can act. A connection that has not joined can no longer send orders, and each guest's request rate is capped.
- A guest can no longer break a shared world. The game refuses oversized names, cable routes, trays and lists from a guest, and trims a world that already holds them when it loads.
- A world too large to send now says so. A guest joining a world over the size limit sees a message saying why.
- A full host tells you it is full. Joining a host that has no room left shows a clear message instead of retrying.
- Guests see Relocate greyed out. The host always refused a guest's move, and the button now shows it.
- A failed part in a switch, router or appliance names who it hits. The incident lists the tenants behind the device and only grows more severe for tenants it actually cuts off.
- Growth requests read the same fit as new prospects. A request that would fill your servers' memory says Tight fit in the warning colour, and one your disks cannot hold says to add storage.
- Service names match everywhere. Every panel uses the game's own name for each service, in the same order.
- Saves from early playtest builds load again. A save that listed the old WAF appliance failed to open, and it now opens with that box as a firewall.
- Dedicated servers load mod scenarios without edits. A mod scenario allows co-op unless it says otherwise, so you no longer edit its file, which broke its Workshop link.
- A broken translation file no longer crashes the game. An override file with bad entries loads its good ones and skips the rest.
- Equal repair-time limits no longer crash the game. A mod or scenario that sets the same minimum and maximum repair time now uses that time.
- Far racks show their latest numbers in multiplayer. A server or switch far from you no longer keeps an old reading after its traffic settles.
- The internet view no longer shows a developer note. A line with no traffic on its router read a paragraph of internal notes after "NOTHING ROUTED HERE". It now reads just that and the router.
- Content packs load safely. A pack scenario nested too deeply is refused with a clear error instead of hanging the game, and a pack's own customer types survive saving and loading.
Notes
- Save compatibility. The save version is now 61. v0.1.80 opens v0.1.79 saves, and installed network cards keep their ports and cables. v0.1.79 cannot open a v0.1.80 save.
- Version lock. The game and its server now speak protocol 76. A v0.1.79 player cannot join a v0.1.80 world, or the reverse.
- Mod API minor 3 breaks script mods. Port numbers are now 16-bit, so the game refuses a minor 2 script mod until it is rebuilt with uptime-mod 0.4.0. Lua mods set `api_minor` to 3.
- Module commands. `BuyModule`, `InstallModule` and `RemoveModule` replace `BuyHostNicModule`, and `BuyModule` with `install` buys and seats a card. Mods get `BuyModule` only. `PickupItem`, `DropAtPosition` and `SellCarriedItem` accept `Module`.
- Diagnosis fields. `CandidateCauseView` gains `entity_kind` and `sub_id`, and `DiagnoseCause` now takes both, so a script mod that sends it must be rebuilt.
- Nested hold goals survive a reload. In a custom or mod scenario, a "hold for N days" win inside an any or all goal keeps the days it has served when you load a save.
- Two lose conditions removed. `DataLossEvent` and `AuditFailure` never fired and are gone from scenario files. A pack naming either is refused by name; every other pack still loads. `SecurityBreach` now ends the run.
- API: revenue by service is a list. `revenue_breakdown` replaces the seven per-service `*_dollars` fields with `services`, one row per service with `class_tag` and `dollars`. A declared break in mod API minor 3.
- API: rack and slot fields. Racks gain `buy_limit_reached`, and chassis module slot rows gain `port_base` and `port_stride`, the slot's first port index and its port spacing.
- API: new fields. The catalog lists every service with both names in `service_kinds`, and growth requests carry `fit_mem_after_pct`.
- Incompatible mods say why. A refused pack names the manifest field at fault, and a missing dependency names the pack to select too. `ModErrorView.field` is now filled for these rows.
- Mod panel and pack limits. A script mod's slider value between two steps shows at the nearest step. Blank panel and config titles are refused, and each hardware file takes at most 128 rows.
- Mod scenario `coop` defaults on. A pack scenario that leaves out `coop` now allows co-op. Set `coop: false` to keep one single-player. `--list-scenarios` now shows pack scenarios with `--mods-dir`.
- Dedicated server token. A dedicated server refuses to start with a `--host-token` shorter than 16 characters.
- Cart commands. `CartAddItem`, `CartSetQty` and `CartRemoveItem` take an optional `fill_module` that fits that card in every NIC slot of a server line.
- HTTP API fields added. Switch and host views gain `module_slots`, and switches gain `no_control_plane` and `line_card_capacity_gbps`. Incidents gain `root_entity_kind`, and `EddView.display_name_plain` is new.
- HTTP API fields that changed meaning. Port indices can exceed 255. `resale_dollars` on servers, switches and carried or loose items includes seated cards. `affected_host_ids` now lists the server for part and card failures.
- Rack and port fields that changed meaning. Rack occupant `power_w` is the live draw, 0 when unplugged, and internet boxes gain it. `PortView.active` stays false on an internet port whose line has no subscription.
- Game snapshot fields. `CatalogSnapshot.modules` (rows are `ModuleSkuView`) replaces `nic_modules`. SKU views gain `slots` with each slot's `port_stride`, and host SKUs gain `nic_fills`.
- Cart, delivery and carry fields. Cart rows gain `fill_module` and `fill_count`, and cart and delivery `capex_dollars` include the cards. Boxed, carried and loose servers gain `fitted_modules`, and carried items gain `module_family`.
- Held cards in every update. Each delta carries `carried_modules`, the network cards in every player's hands.
- New events and incident kind. `ModuleInstalled` and `ModuleRemoved` are new events, `ModuleFailed` is a new incident root kind, and command replies can name a `Module` they created.
- Message arguments. Events, incidents and replies pass device names beside the ids they already carried, so integrations reading ids keep working. A card reply's `slot` argument counts from 1.
Added
- Each availability zone has an Open to customers switch on the Relocate page. A zone you buy starts closed, so customers do not expect a share of it until you open it.
Changed
- Cosmetic mods join a save you already started. A pack that only changes looks or sound applies on your next load. Gameplay packs stay with their original save, and loading names any it left out.
Fixed
Racks and cabling
- Wall racks seat gear on true 1U spacing. Devices no longer clip through the bottom of the cabinet, and racked firewalls sit flush with the rail.
- Firewalls no longer draw a stray square off their back. The power inlets and bezel sit on the chassis itself.
- Picking a device from the back of a rack selects the one you aim at. Before, it took the device one U below, including shelf switches and deep appliances.
- Clicking an empty rack slot works the same as pressing E. Before, a click on the bottom slots did nothing.
- Mounting a rack in a busy colo no longer freezes the game a moment later. The pause ran about half a second.
- Breakout cables on short runs fan out of the connector in a smooth curve. They no longer kink off the trunk.
- Aiming at a full cable manager slot turns the reticle red. The game refuses the clip and says the lane is full, instead of drawing the cable there and moving it when you finish.
- A link aggregation between two switches through a patch panel comes up. The far switch no longer shuts its ports as if the cable formed a loop.
Moving around
- You no longer snag on walls. You slide along interior and exterior walls, and you no longer need Unstuck after brushing one.
- Desks, chairs and the colo mezzanine stairs are solid. You can no longer walk through them, and they block what you can aim at behind them.
- The NOC wall screens can no longer be used from behind the NOC's back wall. Props in front of a screen block it too.
- Setting something down against a wall or a rack keeps it out of the wall or cabinet. It lands in front of you on open floor.
Customers and status
- A switch whose uplink is full now says so, with the uplink's load and limit. It no longer shows Saturated beside a nearly idle backplane.
- A spare switch carrying only standby links no longer makes a tenant read as partitioned. The tenant follows the switch its traffic actually uses.
- Isolating a storage tenant onto fewer hosts than its bucket spans now works. The bucket moves onto the dedicated hosts, and the game warns you that it is under-replicated.
- A service whose control plane has outgrown its host says so on the tenant's card. Before, the card claimed the VMs or database would be placed on the next pass.
- Customers stop complaining about lag once an outage is over. A tenant that normally runs close to its latency budget now clears the alert, and an unhappy customer's mood recovers while their service stays healthy.
- Customers an outage moved onto fewer machines spread back out on their own. When the failed switch, host or zone comes back, the game moves their VMs back, so they can leave probation.
- GPU and mainframe customers only count zones that can run their work. Racking any server in a new zone no longer puts them on probation for a zone they could never be placed in.
- Demand surges count as capacity requests. Auto-approve grants them within the same limit as growth requests, and the Requests page lists every customer request with Approve and Refuse.
Interface and codex
- One controller button can be both Interact and Menu confirm. Rebinding one no longer strips the button from the other, so you can still cable and rack. Buttons stay unique within one kind of action.
- The rack info panel and codex articles stay on screen at larger UI scales. The rack panel also has a close button, so you can shut it on the Steam Deck.
- The codex lists the Everyday Server at 400 W. That matches the shop and what the server draws in game.
- The codex latency budget article states the real alert rule. A warning starts at 75 percent of the budget, and recovery needs p99 under 67.5 percent with failed requests back near normal.
- Tab switches between plain and engineering names only when pressed alone. Ctrl, Alt or Shift with Tab no longer flips it, and shop search finds an item by either name or its SKU.
- The flood protection quest highlights the Flood protection page. It pointed you at Security before.
Notes
- Save compatibility. The save version is now 60. v0.1.79 opens v0.1.78 saves with your zones open, but v0.1.78 cannot open a v0.1.79 save. Items stuck inside a wall or cabinet stay put.
- Load once to heal. Customers an earlier outage left on fewer machines start spreading back out when you load. You see the moves in the migration queue, with no downtime.
- Version lock. The game and its server now speak protocol 75. A v0.1.78 player cannot join a v0.1.79 world, or the reverse.
- Cosmetic mods in multiplayer. A cosmetic pack that joins a hosted world becomes one that joiners need installed.
- HTTP API fields added. `SwitchView.trunk_capacity_gbps` is the switch's uplink capacity, and `saturated` judges that uplink, not the backplane. `RackView.full_strip_bands` lists the cable manager bands that are full. `AzView.open_to_customers` says whether a zone counts for customers.
- The mod API is v1alpha1 minor 2. Mods see the new fields and zones gain `open_to_customers`. Scenario tuning gains `incident_p99_recover_frac`. Mods built for minor 1 keep loading unchanged.
- Script mods run on Wasmtime 49.0.5. It fixes three security advisories. Your mods run the same as before, and saves that use them load without a warning.
Fixed
- Uplinks from a layer 2 top-of-rack pair to layer 3 switches stay up. v0.1.77 joined every rack on one layer 3 switch into a single network and blocked their uplinks as loops.
- Subscribed Workshop mods start downloading when you open Mods. Before, they could sit at Not downloaded. Until the files arrive, the row shows the mod's Workshop title instead of an item number.
- A Workshop row turns into its mod when the download finishes. You no longer close and reopen the Mods panel to see the mod's version and content.
- The mod upload names each step as it starts. It no longer prints "0 of 0 bytes" for every step.
Changed
- The mod upload checks your preview image first. It must be a JPG, PNG or GIF of at most 1 MiB. A wrong file stops the upload before anything reaches Steam.
- The mod SDK folder in the game includes its licence and terms. You get the SDK's MIT licence and the terms for the example art beside the tools.
- The Workshop guide lists the rules for shared mods. It covers what you may upload, how to report a mod, and what uploading lets us do with it.
Notes
- Save compatibility. The save version stays 59, so v0.1.78 saves open in v0.1.77 and v0.1.77 saves open here. Uplinks blocked by v0.1.77 come back the first time you load.
- Version lock. The protocol stays 74, so v0.1.77 and v0.1.78 players and servers can join each other.
Added
Content packs
- Content packs add hardware and scenarios. Hardware brings its own models and icons to the shop. Packs can also add tuning and translations, and replace built-in items. Scenarios join the new run list.
- The Mods panel turns packs on for new runs. It shows what each pack needs and replaces. A pack the game refuses reads "Can't load", with the reason in its details.
- Content packs can ship 3D models for new switches and hosts. They show in racks, on desks and in your hands with working port lights, and the shop preview updates when a model finishes loading.
- The uptime-mod tool builds the light layout from a model file. It warns you about every light it drops, and the SDK includes an example pack with a model.
- A save loads with its own packs. If a gameplay pack is gone, you get a recovered copy that removes its hardware and refunds resale value. A missing art or translation pack loads without it.
Script mods
- Script mods run your own logic in Rust or Lua. They run in a sandbox, and each mod declares the parts of the world it reads.
- Mods can add panels, settings and saved state. A mod's panels appear in the game, its settings show in the Mods panel, and its state is stored in your save.
- Mods can raise toasts. A mod gets two per step, 160 bytes each, and shows one at a time. You can turn off a mod's notifications, and everyone in a multiplayer world sees them.
- Mods can place a HUD card. The HUD shows at most four. Turn off Show on HUD in the Mods panel to hide a mod's card. Everyone in a multiplayer world sees the same cards.
- Mods can declare keys that open their panel or press its buttons. The keys work without the laptop. Rebind or turn them off in Controls, and a mod cannot take a key the game uses.
Workshop and tools
- Mods come from the Steam Workshop. Subscribed mods show up in the Mods panel next to the ones in your local mods folder.
- The uptime-mod tool and its docs ship as ModSDK. It makes, checks, packages and uploads a mod, and the folder includes guides and working examples.
- An example alerts mod shows toasts, a HUD card and keys together. It warns you when an attack starts and counts open incidents, and F6 opens its panel.
Multiplayer
- Multiplayer worlds can use mods. Joiners need only packs that carry art or translations. Other packs reach them through the world, so they install nothing else.
- A missing pack refuses the join and lists what you need. Each pack in the list that came from the Workshop has a Subscribe button.
- Guests can press mod panel controls in a hosted world. Panels on a dedicated server are view-only.
- Dedicated servers load mods. Start one with `--mods-dir` and `--enable-mods`, and give operator settings with `--mod-config`.
Changed
- Modded runs earn no Steam achievements, stats or leaderboard scores. The game tells you so. A mod that only adds art or translations does not count.
- Load balancers bill like an AWS application load balancer. They charge $0.0225 hourly plus $0.008 per GB forwarded, not $0.80 per million requests, so they earn up to 35 percent less, most with small responses.
- Busy worlds send about 43 percent less incident data. The game no longer repeats a confirmed cause, sends shared message details once, and links articles by id, so incident updates stay small.
- Disk failures follow each server's drive count. Three models gained drives and fail them more often; five lost drives, including the Storage 4U and Mainframe 4U, and fail them less.
- Evicting a Reserved tenant early can cost most of its prepay. The refund covers the whole unused term with no cap, so you pay back more of the prepay than before.
Fixed
Hardware and networking
- Eight server models now list the drive count their chassis shows. The Balanced x86 2U lists 8 drives, not 12. Total storage is unchanged, and loading a save corrects the servers you already own.
- Built-in switch, gateway and appliance alarm lamps stay dark until the device is faulted. Fan lamps stay green through a fault, as on real hardware.
- You can bond cables between two MLAG pairs. A paired switch's console lists its partner's cables on a Partner ports tab, and the loop breaker no longer shuts one cable of the bundle.
- Two MLAG pairs joined by more than one link no longer keep a loop open. The game shuts the extra link whole, and a pair whose peer-link fails keeps both switches forwarding.
Customers and services
- A server failure whose customers all moved cleanly no longer rates as a top-severity incident. Their revenue no longer counts as at risk, while a customer whose restart stalls still counts.
- The Incidents list shows IN PROGRESS while a repair is open. It goes back to OPEN if you cancel the repair.
- A security breach incident closes when the breach cooldown ends. Before, it stayed open until the customer left.
- Relocating while a new customer is onboarding no longer keeps their signup bonus. If they leave mid-move before you served them, the game takes back the bonus and prepay, as with any failed onboarding.
- Evicting a Reserved tenant refunds the unused part of its prepaid term. You pay that share back plus the tier penalty, where before the refund stopped at one month of its bill.
- Load balancers that grow now spread their new proxies like every other service. Before, they ignored your spread setting, and the spread check reported them short until you rebalanced by hand.
- The Overselling CPU article now explains what lowering the overcommit dial does. The newest VMs move off hosts that sit over the new ceiling.
Multiplayer and interface
- Clicking a cable port before its details load now says the click missed a cable port. Before, it called the port inactive. The hover hint now names the same port the click picks.
- The multiplayer roster card no longer jumps in height when a player joins or leaves. The card keeps a steady height while the player list changes.
- The game no longer blames another player's link when the host lags. The host now sees that the lag is on their own machine, and the struggling count leaves the host out.
- Hints and highlights and Teaching arcs stay off when you start or load a run. The game now applies both saved switches at the start, so you no longer toggle them again.
- The Migrate menu names zones correctly after a site move. It shows each zone's own name, such as AZ-2, instead of an internal number like AZ 7.
- Picking up a delivered box no longer times out in multiplayer worlds with many open incidents. The cutover countdown no longer jumps, and delivery and migration updates reach every player at least every two seconds.
Notes
- Save compatibility. Save version 59 also stores recent mod toasts. An older build refuses a v0.1.77 save instead of dropping its mods, older saves load here, and loading drops any bad toast.
- Version lock. The game and its server now speak protocol 74, so joiners need the same version. A v0.1.75 client cannot join a v0.1.77 server, or the reverse.
- HTTP API field added. Host and switch views add `model_sku`, the base model a pack item renders as. Empty means `sku_name`, and `docs/openapi.json` shows it.
- Wire-only fields added. The version lock covers `CatalogSnapshot.gameplay_modded_ever` and `.mods`, `WorldSnapshot.mod_panels` and `CommandResult.created`. `ModPanelView.input` reads `players` or `none`.
- New example mods. `com.uptime.example-alerts` and `com.uptime.example-model` join the examples in the SDK folder.
- New snapshot sections. The wire adds `mod_hud`, `mod_keys` and `mod_toasts`, `ModPanelView.point` can read `hud`, and `mod_panels` no longer carries HUD panels. The HTTP API is unchanged by them.
- Load balancer prices reach existing saves. Loading a save moves its load balancers to the new prices. Scenario files now set `per_lcu_hour_microcents`, and the game ignores `per_million_requests_microcents`.
- HTTP API incident readings changed. On `IncidentView`, `revenue_at_risk_per_hour_dollars`, `max_sla_tier`, `compliance_at_risk` and `band` now count only the customers the incident harmed, and `work_order_ids` lists open orders only. Integrations see different numbers.
- HTTP API incident links changed. On `IncidentView` and `CandidateCauseView`, `articles` is now `article_ids`. Titles come from the `article-<id>-title` keys.
- HTTP API incident causes are leaner. `causes` no longer repeats a confirmed root as its first entry, and `symptom_msg_args` is empty when it equals `root_msg_args`.
- Open incidents keep the old reading. Incidents already open in an older save keep their old numbers until they close.
- The mod API is v1alpha1. It may change between releases, and mods name the minor version they were built against.
- The mod API is v1alpha1 minor 1. Mods built for `api_minor` 0 keep loading unless they read incidents or set `per_million_requests_microcents` or `eviction_refund_cap_months`. Those must be rebuilt against the new SDK.
Fixed
Repairs and failures
- Start repair brings a failed server back. It returns minutes after the work order ends, unless a dead part, a tripped breaker or a zone outage holds it down, and #ops says which.
- A repaired server no longer waits for a neighbour. Replacing a dead part brings the server back even while another server in its rack is down with its own, unrelated fault.
- The event feed says why a server failed. It names a hardware fault, a tripped rack breaker, a zone power loss, overheating or a failed part instead of "downstream of upstream failure".
- A zone hit by a second outage stays down until the second one ends. It no longer reads healthy when the first outage ends while some of its servers are still dark.
- Loading a save leaves uplink alarms alone. An open uplink saturation alarm no longer closes on load, so a reload no longer pays out reputation and skill or restarts the alarm's clock.
- Removing one switch of an MLAG pair dissolves the pair. The other switch no longer reports a partner that is gone until you reload.
Customers and capacity
- Customers stop asking for VMs no server on sale can hold. A busy customer asks for the largest VM a server you can buy at your site can run, or does not ask.
- A refused growth request stays refused. The customer no longer asks again every minute and costs you trust each time you say no. They ask again once you install a new server.
- Prospects size their VMs to servers your site can buy. The game caps a prospect's VM size by the servers on sale at your site, not by every server in the catalog.
- GPU and special servers no longer count as room for ordinary VMs. A prospect or growth request that fits only on a GPU, RISC-V or mainframe server now reads as not fitting.
- Passing on a GPU prospect you cannot seat is free. Turning down a GPU customer when you have no free GPU no longer costs reputation.
- Customers lose patience during a long outage. They move from polite to worried, angry and leaving at the pace their contract sets, and their messages show it. Before, their tone never changed.
- Customer trust falls during an outage. A customer's trust now drops through a long outage at the rate their contract sets, and it feeds their mood and whether they stay. Before, it never moved.
- Customer message moods vary again. After about three months of play, every customer message had used the same mood.
- A customer with nothing running shows nothing serving. The customer page no longer keeps the last serving count after a customer loses every unit.
- A storage bucket short on disk asks for storage. A bucket that the failover reserve blocks now says it needs storage, not compute.
- Managed databases count as noisy neighbours. The CPU contention alert names a database when it is the heaviest load on a server, and the server console counts every service's CPU right after a load.
- A refused isolate keeps the customer's segment. If the game refuses to isolate a customer onto dedicated servers, the customer stays on their old segment instead of dropping to the shared one.
Network and traffic
- Traffic between your servers stays off the internet line. During a flood, server-to-server traffic no longer counts against the site's inbound line, so a flood that fits the line no longer fails requests.
- A switch behind an inline firewall or a patch panel can saturate. Its uplink now has a limit, so the switch raises a saturation alert when that uplink fills.
- Targeted floods spread across a customer's zones. A flood aimed at a customer in several zones now hits some and spares others, where it used to hit all or none. A running flood may move.
- Every bonded port counts its traffic. Each member of a server's bonded connection now adds to its lifetime traffic counters, not only the first.
Interface
- Marking ports for a bond no longer resizes the console. Switch, gateway and appliance consoles keep room for the Bond button, so the window keeps its size as you mark ports or change mode.
- The rack side panel stays on screen. A full rack, a long name or a saturated switch no longer pushes it off the right edge, and long readouts show in full on hover.
- CPU figures use one word for each thing. Physical cores read pCPU on the Capacity page and a prospect's free capacity, as on the server console, and vCPU is the overcommitted amount you can sell.
- Capacity rack space skips gear in your hands. The Capacity page no longer counts a server you carry as used rack space.
- Gateway shop rows show their ports. In engineer mode a gateway lists its port summary as its name, as switches and appliances do.
- Four port types get proper labels. SFP56, SFP112, QSFP+ and QSFP56 ports show readable names instead of raw ids.
- Sandbox settings stay in their shown ranges. The game holds a value set during a run, from the host or the API, to the range the configurator shows.
- Hints and the Codex match the incident cards. The text that tells you what to check reads the same in all three places.
- More of the game can be translated. The security log, the spread warning and the sandbox labels now use translatable text. The English wording stays the same.
Notes
- Saves load both ways. A v0.1.75 save opens in v0.1.74, and older saves open in v0.1.75.
- Version lock. The game and its server now speak protocol 73, so a v0.1.75 client cannot join a v0.1.74 server or the reverse.
- Load once to heal old id counters. A very old save that lacks some id counters no longer hands out an id it already uses. Nothing moves and nothing is renamed.
- Finished work orders are cleared. The game removes a work order a day of game time after it closes, so a long game keeps a smaller save.
- API fields removed. About 60 view fields that nothing read are gone, including the host monthly cost fields, cable routing fields and customer lifetime request counts. `docs/protocol-history.md` lists each.
- API fields changed. `SiemEventView` carries `msg_key` and `msg_args` instead of English text, `ServiceDemandView` adds `reason_key` and `reason_args`, `HostSkuView.arch` and `accelerator` use the enum name, and `HostUplinkView.lag_mode` is typed.
Fixed
Network and failures
- A busy switch only fails traffic that crosses it. A saturated or storming spine no longer drops requests routed through a different spine, and MAC flap losses follow the path each flow takes.
- Workloads leave a failing zone for servers that stay up. When a zone, rack breaker or heat runaway takes servers down, their workloads no longer move onto a neighbour that fails seconds later.
- CPU contention incidents point at the busiest server. A customer spread over several overloaded servers is sent to the one under the most load, not the lowest numbered.
- Cutting a carrier line takes workloads off stranded servers at once. Servers that lose their only path stop serving right away, and the outage counts against their customers.
Customers and capacity
- Servers in maintenance no longer count as free room. "Can you serve it", growth requests, the capacity meters, segment capacity and the GPU verdict leave them out, so a prospect that fits can be placed.
Multiplayer
- Big dedicated servers no longer stutter when a server fails. With thousands of servers, every failure or recovery froze the world for about half a second. The game now handles it in a few milliseconds.
- Multiplayer sends far less when a message arrives. On big worlds, each new message or read thread sent every player the whole comms list again. Now the server sends only the channels that changed.
Interface
- The Relocate button moves you again. Since v0.1.72 the game refused every one-press move, and only a schedule picked in Options worked. The button now moves you on the standard schedule.
- The comms panel loads messages on big worlds again. With many customers, threads stayed empty and Mark all read did nothing. Both now update within a few seconds.
- Rack lamps match the lamps in the world. A switch, router or firewall in the rack view shows the same lamp as the device for storms, MAC flaps, saturated fabric and dead power supplies.
- Spread advice gives you numbers. A stacked service's card now says how many sit on one machine and how many to move, using the same per-machine limit as the spread warning.
- Placement problems read the same everywhere. The customer inspector and the repair hints give each reason in the same words, and a cross-network outage names the customer instead of a number.
Notes
- Saves load both ways. A v0.1.74 save opens in v0.1.73, and older saves open in v0.1.74.
- Version lock. The game and its server now speak protocol 72, so a v0.1.74 client cannot join a v0.1.73 server or the reverse.
- The metrics API now lists more series. Gateways list 7 series, firewalls 10 and SIEMs 12; fleet adds reputation, drops, scrubbing, events and attack series. Log series now answer only on a SIEM.
Fixed
Customers and capacity
- An Everyone pool leaves VRF tenants on their own servers. Adding a server to it, or switching back to Everyone, no longer fails on capacity or moves a VRF's workloads off its dedicated servers.
- "Can you serve it" counts only servers a prospect can use. Servers dedicated to other customers or pools no longer make a prospect, or a customer's growth request, look like it fits.
- Picking a VRF judges the prospect against that VRF. The verdict, the Accept button and a new GPU row read the servers that VRF can use, including a VRF with no servers of its own.
- Approving one more load balancer adds one. A load balancer request now counts whole load balancers up to the cap of four, and a customer at the cap no longer asks.
- CDN caches show the CPU they use. Each cache reserves 1 vCPU and 512 MB, and its row shows its real draw, so a busy server no longer shows load while its rows read zero.
Hardware
- Background internet noise skips servers in maintenance. The gateway firewall drops the noise it defends against, and anything it lets through lands on a working server, never one in maintenance or switched off.
- Network jacks no longer read "AC inlet". Servers moved to a bigger memory model in v0.1.64 kept their old port layout, so two network jacks showed as power inlets.
- A cable is placed in the colour you saw. Finishing a run on a slower port no longer swaps the cable to another colour as it drops to the slower cable type.
- The rack panel shows the power the rack draws now. It follows load and wear like the ops console and the breaker, instead of adding up each server's rated maximum.
Interface
- Each host pool has a Release button again. It sits beside Edit on the laptop's Isolation page and opens every server in that pool in one press.
- Options on the move-site page always opens. It works whenever the Relocate button is lit, instead of doing nothing after a gate dips for a moment.
Notes
- Saves load both ways. A v0.1.73 save opens in v0.1.72, which ignores the two new CDN reservation settings, and older saves open in v0.1.73.
- Load once to heal. Loading puts every server's ports back in its model's layout, books each CDN cache's new reservation, and removes any cable saved with one end unplugged.
- Loading can unplug a cable. A server saved with more network ports than its model has loses the extras, keeping its cabled ports first, and a cable on a removed port comes off.
- Version lock. The game and its server now speak protocol 71, so a v0.1.73 client cannot join a v0.1.72 server or the reverse.
- New API fields. Network segment views gain `dedicated_free_vgpu_slices` and `dedicated_total_vgpu_slices`, the GPU slices on a VRF's dedicated servers.
- Gateway LAN egress counts only forwarded traffic. `lan_egress_gbps` now reports what the gateway passes to your servers, so with the firewall on it excludes background noise the firewall drops.
- API removals. The `UnplugCableEnd` and `ResumeDanglingCable` commands and the cable views' `dangling_end` field are gone; nothing in the game sent or set them.
- Prospect fit fields changed meaning. Prospect and growth request `fit_*`, and growth `segment_*` meters, now count only servers that customer can use, per service. Prospects gain `landing_*` capacity fields and `segment_fits`, one verdict per VRF.
Changed
- Four customer types now want a SIEM before they sign. Enterprise pilots, government contracts, fintech and healthtech startups need a powered, cabled SIEM on your network. Certification demands stay off until you can earn certificates.
- An unbonded server uses one network card. The second card is a standby that takes over if the first fails. Bond the cards to use both.
- Quotes match the bill. Prospect cards, relocation and margin figures now include the $75, $200 or $500 monthly platform fee.
- Down capacity stops billing, and the spend cap covers more services. Compute, storage and requests stop billing while their server is down, and CDN, object storage and load balancers stop serving at the spend cap.
- Emergency bailouts scale with your costs. Each one covers at least 30 days of your monthly costs, never less than $25,000, and still grows 1.5x per use, three per run.
- "Cover the edge" asks for the right flood protection. It asks for the carrier's Proxy plan at a garage or basement, and the Basic Scrub router add-on at larger sites.
- Hints and teaching arcs are a world setting. In multiplayer only the host, or a dedicated server's admin, can change them.
Fixed
Performance
- Large outages, busy networks and big fleets no longer slow the game. Site-wide incidents, filling links and quest tracking all keep the game at speed as your operation grows.
Hardware and networking
- Carried and unplugged gear behaves like it is off the rack. It adds no heat, survives its old rack's failures, and gets the same depth and breaker checks when you rack it again.
- Selling and recalling gear leaves nothing behind. Routers drop their add-ons and cables cleanly, a chassis drops its line cards, and your lists keep their order. A slot past a rack's top is refused.
- Wall-mounted routers work like racked ones. You can put one back on the wall, and attacks on it count toward its zone.
- Bond changes apply at once and keep your settings. Mode, minimum links and WAN policy reroute traffic on the next tick, and a refused or repeated LACP request no longer loses the bond.
- Bonds count only real data links. Keepalive and stack cables cannot join, breakout legs can, a patched member counts at its slower end, and pulling the last cable clears the bond.
- Link states read true. A link stops reading saturated when traffic ends, an upgraded line stays down through an outage, and a split MLAG pair drops half the requests crossing either switch.
- An overheating one-server room can lose its server. Thermal runaway now fails the only server instead of raising an alarm with no effect.
Customers and money
- Changing a server's job queues its workloads. They copy over the network through the migration queue like any other move, and the confirmation counts the moves queued.
- VMs keep their spread when they move. Drains, failures and rebalances count a customer's VMs per zone and never stack two on one empty server, and the move warning matches the spread advisory.
- Moves and resizes respect real room. A resize no longer overfills a server, and VMs leave a server that is unwired and powered off.
- Refunds and fees follow the contract. A Reserved customer who fails onboarding gets the prepay back, evicting an unbilled customer costs its contracted fee, and a bailout onto debt resets its deadline.
- Margins and part orders add up. Roster margins include Functions provisioned concurrency, and moving site keeps the books for carried part orders and cancels swaps on parts that left.
- Customers explain why they leave. The log names missed availability, undelivered services, slow storage or broken spread instead of a health number.
- Requests and prospects fit your fleet. The growth card agrees with Approve for resize requests, and a prospect who signs while you have one server asks for a single service.
- Boxes stay where you put them. A pallet keeps its stacking order, and moving site brings your unopened boxes and part orders, with arrived spare parts usable again.
Progress
- Goals and scenario rules count what they should. The racked-servers goal counts only racked servers, arch-restricted scenarios refuse the wrong CPU, and Keep Playing needs a win first and the host's say.
- The Achievements panel lists only achievements you can earn. Four that have no way to unlock yet are hidden until their content arrives.
Multiplayer
- The shared world keeps real time. It ran about 4% slow before, because the host's clock lost a few milliseconds on every tick.
- A busy host no longer falls into a lag spiral. Each tick now hears every player before it moves on, so players stop drifting further and further behind on a large world.
- Joining and reconnecting tell you what happened. The loading screen keeps drawing while Steam verifies you, a final refusal stops reconnects and says why, and you are back only when the host confirms.
- Unfinished joins cannot lock players out. They take no seat, leave difficulty alone, never unpause or save the server, and are disconnected. One Steam account cannot fill the join queue.
- Difficulty matches your group from the first tick. Building a world or loading a save with friends connected scales demand to the whole group.
- A dedicated server resists a misbehaving client. It caps each player's position updates and spaces its empty-server saves, so one client cannot flood the others or overwrite every backup.
Incidents
- Incident cards name the right cause and customers. DDoS cards list only the attacked zone's customers, breaches follow the target's own edge, and "Servers can't see each other" covers switches joined only through the router.
- Incident figures stay current. Revenue at risk includes Functions warm capacity but not the flat platform fee, and a latency incident closes once the customer's traffic stops.
Interface
- Panels show the right numbers. Host CPU rows add up, switch charts plot real speeds, light links read "<0.1%", reopened charts keep their data, and the Internet screen and wallboard agree on WAN traffic.
- Text fits and warnings say what they mean. The capacity wall and shop descriptions no longer cut off, and the cooling warning reports heat against cooling capacity instead of naming breakers.
- Consoles and prompts stay in step. Guidance shows when you open a console, new leads are announced by who they are, a diagnosis shows only its latest result, and Q closes the robot console.
Notes
- Saves load both ways. A v0.1.72 save still opens in v0.1.71, and older saves open in v0.1.72, including ones with queued function calls, which load empty.
- Version lock. The game and its server now speak protocol 70, so a v0.1.72 client cannot join a v0.1.71 server or the reverse.
- Steam dedicated servers answer nothing before an approved join. A connection must pass the Steam ticket check before the server accepts any other request from it.
- Load once to heal. Loading clears workloads on sold servers, add-ons left by sold routers, bad bonds and stranded spare parts, and restores settings a difficulty change undid. Nothing moves.
- Load once to tidy older saves. Absorbed debt loses its old deadline, invalid numbers reset, guest-saved toggles and stale trolley entries go, bonds left by a dropped switch pair clear, and achievements follow the build.
- Churn reasons are cause codes. A probation churn event's `reason` is now `health-availability`, `health-unmet`, `health-storage`, `health-spread` or `health`, not a health number.
- API removals. `/api/v1/work-orders` and `/api/v1/work-orders/{id}` are gone, as are the `Diagnostics` fields `active_work_orders`, `pending_bond_requests`, `proposed_bond_groups` and the `uptime_work_orders_open` gauge.
- Prospect revenue estimates include the platform fee. `estimated_monthly_revenue_dollars` reads $75 (Silver), $200 (Gold) or $500 (Platinum) higher than in v0.1.71.
- Stricter command checks. The game refuses an unknown relocation window or NIC team mode, NaN or infinite numbers (`invalid_number`), a 0 m cable and a spread setting for an unknown customer.
Current Release
Build 25957697
Uploaded Oct 16, 2026
Requests are paused right now
PC Games requests are paused for now while we work through the backlog. Other areas and voting stay open.
Login Required
You need to be logged in to request game updates. Would you like to go to the login page?
Request Update
Request an update for Uptime: A Cloud Provider Sim? This will notify uploaders that this game has an available update.
Request Submitted
Your update request has been submitted successfully! Uploaders will be notified.
Request Failed
Failed to submit update request:
System Requirements
How to Install
Uptime.exe to play
Troubleshooting tips
• Run Redist/_CommonRedist installers if game won't start
• Add folder to Windows Defender exclusions
• Run as administrator
Download
Direct link available
Needed when extracting (WinRAR / 7-Zip) — not a site login.
61D839CA
f341010387672b0a4e5a1f6cb070ea34c2b4c7d036f8edf65ae7092f2d88fc77
You May Also Like
View AllDownload Uptime: A Cloud Provider Sim for PC with a direct link or via torrent. Get the full version of Uptime: A Cloud Provider Sim for free. Uptime: A Cloud Provider Sim is a Indie released by RubyRack Games.