Hello,
My RS 12000 G12 (Debian 13, kernel 6.12, ext4 on vda ~3 TB, only 20% used) has had persistent, very high disk latency for days, even under light load. My self-hosted Sentry (Postgres/ClickHouse/Kafka) keeps stalling and is currently down.
iostat -x 5 from inside the VM (scheduler=none, steal ~0%, CPU mostly idle):
- w_await: 150–580 ms (e.g. 152.9, 251.4, 254.7, 581.0, 329.4 ms)
- f_await (flush/fsync): 27–266 ms
- %util: 54–78% at only ~30–70 write IOPS (~1 MB/s)
- iowait: 8–19%
What I already tried:
- Checked the guest side: no discard/atime issues, scheduler already "none", I/O pressure is not from CPU or RAM.
- Ran storage optimization in the SCP [confirm result: finished / stuck]. Latency is unchanged and Sentry is still down.
- Cleaned ~150 GB of unused Docker build cache and reduced write load.
The server also restarted on 2026-10-23 ~06:34; the kernel then logged "EXT4-fs (vda4): orphan cleanup on readonly fs" before remounting r/w, so the previous session likely did not end cleanly.
I found a similar thread (RS8000 G11, high iowait) that was resolved by moving the server to another node, so this looks like a host-side storage problem.
Has anyone seen the same on RS G12 recently, and did a node migration fix it for you? I have also opened a support ticket requesting a migration to another host node. netcup staff, if you can have a look at the storage node I'm on, I'd really appreciate it.
Thanks!