I can confirm that the aggressive active MGR memory growth on stops when the map cache is disabled. The active MGR still shows a slow upward trend. Example values taken from top RES (in kb) and the time: 298828 09:48 427224 18:48 491624 12:47 522716 23:12 Even if there is still a small leak it is quite manageable, although I would keep an eye on it for extended runtimes. I don't know whether it was there in Tentacle. The standby MGRs stabilised at around 212500. Regards, Chris On 30/10/2026 18:08, Chris Palmer wrote:
I've just tried that, and it has made an enormous improvement. Both active and standby MGRs are still gently tracking upwards, but they haven't been running long so it may just be natural settling (I restarted each of the MGRs). I've got a note of their sizes and will check again tomorrow.
FYI: this cluster is just a very small one with 18 OSDs. I have no idea whether the leak rate relates to the number of OSDs, PGs, etc.
Thanks so much for the help. Will confirm in a day or two how the sizes are going.
Regards, Chris
On 30/10/2026 17:00, Joshua Blanch wrote:
A new feature in the umbrella for the MGR might be causing the issue.
https://docs.ceph.com/en/umbrella/mgr/administrator/#performance-and-scalabi...
To disable the cache run: ceph config set mgr mgr_map_cache_enabled false
Could you try setting this and restarting the MGR to see if memory growth stops?
- Joshua
On Fri, 30 Oct 2026 at 08:55, Michael Sudnick <ceph-users@ceph.io> wrote:
I encountered the same leak and worked around it by disabling the dashboard, grafana and prometheus mgr modules. Not sure if all are necessary to disable.
-Michael Sudnick
On Friday, October 30th, 2026 at 11:51 AM, Chris Palmer <ceph-users@ceph.io> wrote:
> I've just upgraded a small production light-usage cephadm cluster from > Tentacle to Umbrella. The active MGR now seems to be growing by about > 1GB per hour though. I can't see anything wrong in the mgr logs. > > The only other sign of trouble is that a lot of metrics are not > displaying in Grafana (mainly OSD-related I think - haven't checked it > thoroughly yet). I don't know if that is related to the leak. > > I just had to fail the MGR to get some memory back. It immediately > released about 2.5GB, but is still tracking up gently while in standby. > > Help... (Just adding a large amount of swap...) > > Thanks, Chris > > _______________________________________________ > ceph-users mailing list -- ceph-users@ceph.io > To unsubscribe send an email to ceph-users-leave@ceph.io > _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
_______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io