ServerCrate/Trust/Incidents

Incident history.

When something breaks on a service people pay for, we write down what happened, what you would have seen, why it happened and what we changed. Every write-up stays here permanently. Live status is at status.servercrate.net.

2incidents written up in 2026
0bytes of customer data lost
New backup vaultsResolved the same day

Vault creation failed for some new signups

What customers saw

A small number of accounts that signed up that day saw their vault setup fail. Existing vaults and their backups were not affected.

Root cause

The fast storage pool that holds vault containers had all of its space reserved by thick-provisioned volumes, even though most of it was physically free. New containers had nowhere to be created.

What we changed

Released the unused reservations and switched the pool to thin provisioning, which restored new vault creation. Capacity monitoring now tracks real pool usage instead of reserved space, so this shows up as a warning long before it blocks anyone.

Data impact

None. No stored data was lost or exposed.

Backup vault endpointResolved the same day

A shared rate limit returned temporary errors under load

What customers saw

While one large backup was running, other backup requests could get a temporary HTTP 503 response. Restic retries these automatically. Our log review found the errors only on our own internal test vaults; no customer backup failed.

Root cause

The connection and request limits on our vault edge were keyed on a name every vault shared, so all vaults drew from one small bucket instead of one bucket each.

How we found it

Our restore drill, which we had just grown to a realistic 256 MiB corpus, failed its backup step. The smaller 15-minute canary finished too quickly to hit the limit.

What we changed

Limits are now per vault (16 connections and 60 requests per second each). Every configuration change on the edge now confirms the web server process actually restarted, not only that the file is valid.

Data impact

None. Every affected vault kept healthy, verified snapshots throughout.

Something you think belongs here? Open a ticket from the portal or email hello@servercrate.net and our team will look into it.