search

microsoft / omi

Open Management Infrastructure
Other
360 stars 114 forks source link

omiagent segfault libcontainer.so after upgrading to docker-ce 25.x #755

Closed paltryeffort closed 5 months ago

paltryeffort commented 5 months ago

After upgrading docker-ce from version 24 to 25 on my Azure VM (CentOS 7.9.2009) the omiagent crashes. This creates a crash dump file in /var/opt/omi/run/core.* (5.3M) every minute and eventually fills up the root file-system potentially killing the system. (5MB x 60 x 24 = 7.2GB per day)

journalctl -k -n1 --no-pager

Jan 26 13:43:19 myhostname1 kernel: omiagent[6848]: segfault at 28 ip 00007fdf529fd310 sp 00007fff51bdf490 error 4 in libcontainer.so[7fdf529a0000+8e000]

Details:

# rpm -qa 'docker-ce*'
docker-ce-25.0.1-1.el7.x86_64
docker-ce-rootless-extras-25.0.1-1.el7.x86_64
docker-ce-cli-25.0.1-1.el7.x86_64
# rpm -qf /opt/microsoft/docker-cimprov/lib/libcontainer.so
docker-cimprov-1.0.0-39.x86_64
# rpm -qf /opt/omi/lib/libcontainer.so
docker-cimprov-1.0.0-39.x86_64
paltryeffort commented 5 months ago

Apparently the omsagent does not update when doing a "yum update" on the VM. Why this package is not part of it is beyond me. I would have to go to the Azure portal and enable "automatic upgrade" on the extension "OmsAgentForLinux". This updates the omsagent to a new version:

# rpm -qa omsagent
omsagent-1.17.2-0.x86_64
# rpm -qf /opt/microsoft/docker-cimprov/lib/libcontainer.so
docker-cimprov-1.0.0-43.x86_64
# rpm -qf /opt/omi/lib/libcontainer.so
docker-cimprov-1.0.0-43.x86_64

The segfaults are gone in this version.