Feature #55474
openBLUESTORE_FRAGMENTATION health check doesn't work
0%
Description
It seems like there is documentation around `BLUESTORE_FRAGMENTATION` in the docs:
https://docs.ceph.com/en/pacific/rados/operations/health-checks/#bluestore-fragmentation
But, there is no actual code that triggers this alert, and this is a super critical issue IMHO, since if your fragmentation gets high, it will kill your OSDs with absolutely no heads up or expectations.
Updated by Igor Fedotov almost 2 years ago
- Tracker changed from Bug to Feature
IMO this is a feature request.
Updated by Tobias Urdin almost 2 years ago
What do you even do in that scenario, redeploy the OSD?
Updated by Mohammed Naser almost 2 years ago
Igor Fedotov wrote:
IMO this is a feature request.
Shouldn't we then remove it from the documentation since it lists it as a "health check". The document says "This page lists the health checks that are raised by the monitor and manager daemons." -- nothing currently raises that health check.
Updated by Mohammed Naser almost 2 years ago
Tobias Urdin wrote:
What do you even do in that scenario, redeploy the OSD?
No other choice than this right now sadly, but that's the only best option to do. You'll be in for a very bad `enospc bluefs` assert and OSD crashing/not starting.
Updated by Tobias Urdin almost 2 years ago
Mohammed Naser wrote:
Tobias Urdin wrote:
What do you even do in that scenario, redeploy the OSD?
No other choice than this right now sadly, but that's the only best option to do. You'll be in for a very bad `enospc bluefs` assert and OSD crashing/not starting.
Interesting, this surely seems like a major thing that one should monitor then.