Bug #8648
closedStandby MDS leaks memory over time
0%
Description
I've discovered in my Ceph cluster that the MDS overtime will leak memory. In my case it usually takes a week two or to notice and after two months eventually the machine will kill the MDS server (OOM killed). It looks like this only happens on the standby (stanby-active) MDS, I haven't observed such leaks on the primary. If I switch which one is the standby MDS (due to a version bounce, or restart due to OS fixes) then that machine's MDS starts leaking.
I'm currently running Firefly but this issue has been present for a long time (Dumping & Emperor).
Updated by Greg Farnum almost 10 years ago
I believe we're leaking CInodes in open_root_inode et al.
Updated by Sage Weil over 9 years ago
- Project changed from Ceph to CephFS
- Category deleted (
1)
Updated by Sage Weil over 9 years ago
- Status changed from New to 12
- Priority changed from Normal to High
- Source changed from other to Community (dev)
Any change you can run one of these in standby under massif for a while? that will tell us what is leaking!
Updated by Zheng Yan over 9 years ago
- Status changed from 12 to Fix Under Review
Updated by Zheng Yan over 9 years ago
- Status changed from Fix Under Review to Resolved
fixed by commit eae88dad4c32e4bb5fb255ec4bf1be18b09d498e