<div dir="ltr"><div>I happened to find, that whenever the service is in the "bad state" so it</div><div>will loose the containers on restart it also misses this section in</div><div>"systemctl status"</div><div> Tasks: 19 (limit: 32768)</div><div> Memory: 7.7M</div><div> CPU: 89ms</div><div><br></div><div>I compared to other systemd tools looking at this data:</div><div>- systemd-cgls matches the control group output and is good in all cases.</div><div> The missing entries are more for accounting it seems.</div><div><br></div><div>- systemd-cgtop reports the number of tasks and memory even in the bad </div><div> state where "systemctl status" does not</div><div><br></div><div>Looking further from there I found more by comparing</div><div> $ systemctl show libvirtd</div><div> </div><div>Comparison</div><div>1. exec start lists pid as 0 in bad case</div><div>2. Task/CPU/Mem tracking is near infinit values in bad case (unset?)</div><div><br></div><div>The pid listed in the good case his it's own PID, the bad case has 0.</div><div>MainPID=16252</div><div>[...]</div><div>ExecMainPID=16252</div><div>[...]</div><div>ExecStart ... pid=16252 <- here the bad case has 0</div><div><br></div><div>Note: analyzing that is already in the script I linked in the initial report. So no update for that on this bug update.</div><div><br></div><div>I can't reproduce otherwise than uninstall/reinstall of the package yet.<br></div><div>But there the service is re-started by the maintainer scripts - checking if I can find anything there to trim down the repro narrowing down what is happening.</div>
</div>