Server uptime: what 99.9% actually means
Uptime percentages turned into real minutes per month, what counts as downtime and what does not, and the difference between the network, the machine and your own server process.
6 min read · Updated Oct 21, 2026 · By GuavaHost Team
The percentages, in minutes
An uptime percentage is only meaningful once you convert it. Here is what each one allows over a 30 day month.
| Uptime | Allowed downtime per month | Per year |
|---|---|---|
| 99% | about 7 hours 12 minutes | about 3 days 15 hours |
| 99.5% | about 3 hours 36 minutes | about 1 day 20 hours |
| 99.9% | about 43 minutes | about 8 hours 46 minutes |
| 99.95% | about 22 minutes | about 4 hours 23 minutes |
| 99.99% | about 4 minutes 20 seconds | about 53 minutes |
Two things jump out. First, 99% sounds respectable and is actually terrible: that is an entire evening offline every month. Second, each extra nine costs ten times as much to deliver, which is why 99.9% is the standard commitment for game hosting and 99.99% belongs to systems with full redundancy and a price to match.
For context, 43 minutes a month is roughly one short outage, or a handful of brief blips. It is not "never offline", and no honest host claims that.
Three different things can be down
This is the part that matters most in practice, because when your server is offline, the cause is usually not where people assume.
1. The network. The route between your players and the data centre: the uplink, the routing, the DDoS filtering. When this is down, every server on that network is unreachable at once, and so is the panel for servers in that location. This is rare and it is what uptime commitments are mostly about.
2. The machine your server runs on. The physical node: its CPU, memory, disks and the software that runs the servers on it. A node problem takes out the servers sharing it, while everything else stays up. Hosts reduce this with monitoring, spare capacity and the ability to move a server to another node.
3. Your own server process. The game itself: it crashed, it ran out of memory, a mod threw an exception on startup, you stopped it, or a config change stopped it booting. The network is fine, the machine is fine, your game is not running.
This third category is by far the most common reason a server is offline, and it is the one no uptime percentage covers, because nothing outside your server failed. The useful guides here are the server is stuck starting or offline, reading logs and crash reports, signs you are running out of RAM and server exited with a code on start.
A quick way to tell which one you are in: open the panel. If the panel loads and your server shows Offline, that is category 3 and the console will say why. If the panel loads and the server says Online but nobody can connect, check the address and the port first, because that is usually not downtime at all. Finding your server IP and port and the friends cannot connect checklist cover it.
What usually does not count as downtime
An uptime commitment covers the host's infrastructure. Typically outside it, and this is consistent across the industry:
- Your game server crashing, or failing to start after a change you made.
- A mod, plugin or modpack that is broken or incompatible.
- Running out of memory because the plan is too small for what you installed.
- Restarts you scheduled, and restarts you triggered.
- A game update that breaks your server until you update it.
- Being suspended for non-payment. See missing a payment.
- Problems on your own connection or a player's.
- Scheduled maintenance that was announced in advance.
GuavaHost publishes a 99.9% uptime commitment, and the exact wording, what it covers and how it is applied are on the SLA page rather than in this guide, because that page is the authoritative version and this one is an explainer.
Scheduled maintenance is a good sign
People sometimes read a maintenance notice as a host being unreliable. It is the opposite. Hardware needs firmware updates and panel software needs security patches, and a host that announces a window, does the work and tells you afterwards is maintaining its fleet. The alternative is an unpatched machine that fails on its own schedule instead of yours. What to look for is the handling: notice in advance and a short window.
Making your own server the reliable part
Since your own process is the most likely thing to go down, most of the improvement available to you is here.
- Schedule a daily or nightly restart. It clears memory creep and accumulated entities, and a planned 40 second restart at 5am is better than an unplanned crash at peak time. See scheduling automatic restarts.
- Do not run at the edge of your memory. A server that just fits will eventually not fit. How much RAM you need sizes it, and you can change plan any time with Manage plan; see upgrading your plan.
- Keep backups, and know how to restore one. A restore you have practised is a ten minute outage. One you have not is an evening. See backups and building a backup plan.
- Test updates before applying them to the live world. Game and modpack updates are the most common self-inflicted outage.
- Watch the trend, not just the moment. The Usage card's history view shows memory and CPU over time, which is where a slow problem shows up first. See reading the resource graphs.
- Let the crash tell you. When a server dies, the console's last lines contain the reason, and the panel's Activity log records what happened around it. See the activity log.
Comparing hosts on uptime
A percentage on a landing page is a claim, not evidence. More useful signals:
- Is there a real SLA document, with scope and process spelled out, or just a number in a feature list?
- Is there a status page or an incident channel? A host that publishes its incidents is a host that admits them.
- How do they handle a bad node? Can a server be moved to other hardware, and how quickly?
- Is DDoS protection included and always on, rather than switched on after the first attack? See how DDoS protection works.
- Is support reachable when things break? The GuavaHost team is online 24/7 in Discord, which matters more at 2am than any percentage does.
- Can you test it? A money-back window lets you run your own world on the hardware. How to choose a host is the full checklist, and the 48-hour money-back guarantee covers ours.
Compare that against the realistic alternative: a server on a PC at home, which is offline whenever the machine sleeps, Windows updates or the power flickers. Hosting at home versus hosting with a provider does the arithmetic.
Common questions
How many minutes a month is 99.9% uptime?
About 43 minutes over a 30 day month, which works out to roughly 8 hours 46 minutes a year. It is a realistic commitment for game hosting rather than a promise of never being offline, and it is the industry standard level.
My server crashed last night, does that count against uptime?
Normally no, because the network and the machine were both working and it was your game process that stopped. That sounds like a technicality but it is the right line: a host cannot be responsible for a mod that throws an exception. The console log will say what happened, and a scheduled nightly restart prevents a good share of these.
What is the difference between uptime and an SLA?
Uptime is the measurement, and an SLA is the written commitment about it, including what counts, how it is measured and what happens if it is missed. Always read the document rather than the headline number, since the scope is where the real meaning sits.
Does a restart count as downtime?
A restart you scheduled or triggered does not, and it is also short enough that it rarely matters: most game servers are back in under a minute. That is exactly why a planned restart overnight is a better deal than an unplanned crash during peak hours.
Still stuck? Jump into our Discord at discord.gg/GuavaHost and the team will help you out.
- uptime
- sla
- downtime
- reliability
- hosting
Still need help? Our team is online 24/7 to answer questions.