Bug #58855
cephadm is not setting nofiles and proc limits for docker
0%
Description
Currently only systemd units have LimitNOFILE and LimitNPROC set, but it does not reflect docker.
In addition to that, on modern systems (EL9 - like RHEL9, CentOS Stream 9 or Rocky Linux 9) the default open files limit has been bumped to around 1 bilion (https://access.redhat.com/solutions/1479623 or https://github.com/systemd/systemd/commit/a8b627aaed409a15260c25988970c795bf963812).
Ceph deployed on EL9 is behaving weird with timeouts to any commands after deploying a couple of services (mons, mgrs and a couple of OSDs).
ulimit output on Quincy deployed on Rocky Linux 9 in Docker:
[root@michal-koala /]# ulimit -n
1073741816
[root@michal-koala /]# ulimit -u
unlimited
[root@michal-koala mon.michal-koala]# cat /etc/systemd/system/ceph-47cee5fc-b454-11ed-8871-fa163ee02ea8@.service | grep Limit
LimitNOFILE=1048576
LimitNPROC=1048576
Related issues
History
#1 Updated by Redouane Kachach Elhichou about 1 year ago
- Status changed from New to In Progress
- Pull request ID set to 50200
#2 Updated by Adam King about 1 year ago
- Status changed from In Progress to Pending Backport
#3 Updated by Backport Bot about 1 year ago
- Copied to Backport #58982: quincy: cephadm is not setting nofiles and proc limits for docker added
#4 Updated by Backport Bot about 1 year ago
- Tags set to backport_processed