The price per terabyte and the fact that there are no egress fees. On paper, that is a very attractive deal for backups and container images. It is S3-compatible, so standard tools (aws cli, rclone, Docker registry) connect without changes, and Object Lock is available for immutable backups. That is honestly where the good part ends: cheap and compatible, but not something you can rely on when response time matters. Review collected by and hosted on G2.com.
Performance is unpredictable to the point of being unusable for anything interactive.
We measured it over weeks, not a single bad day:
- median request time is fine (0.1-0.2 s), but p95 is 7-17 seconds and the worst requests take 60-90 seconds;
- on weekday mornings 22-62% of all requests take longer than 2 seconds;
- connect + TLS handshake is under 170 ms, and the same latency shows from three different networks, so the delay is entirely on Contabo's side (time to first byte);
- we get 503 "Slow Down" errors while the rate-limit headers in the same response say 249 of 250 requests per second are still available, so throttling triggers far below the published limit.
In practice this means container image pulls that hang for close to a minute, CI jobs timing out, and backups that cannot be trusted to finish on time.
Support: in mid-October they told us it was a "known issue" that would be "resolved very shortly". It was not. Our ticket has been "under review" since then, with no root cause, no ETA and no credit, while the service is billed in full every month.
Access control is primitive as well: one credential has access to everything in the account, with no per-bucket keys, so you cannot give a backup job least-privilege access.
Fine for cold archives you rarely touch. Avoid it for registries, CI or anything a user waits on. Review collected by and hosted on G2.com.
