Description of problem: debugging crashed ceph deamons in a container is manual and labor intensive Expected results: This includes supporting collection of logs from crashed OSD daemons, with SOSreport running on the host alone, not in every container.
Greg, are you really expecting this for 3.0? I don't even know how SOSreport works, I need to look into it.
Yes please. IF we can't let's at least have a solid plan for what it'd take in 3.1
I just had a look in SOSReport, it has a builtin docker module that is being executed, so to gather logs we could simply let sosreport run by default (it will collect logs from journald, which is where container logs are stored). Or we can be more specific like this: [root@osd0 tmp]# sosreport -v -k docker.all=on -k docker.logs=on I've looked into the generated report and I could successfully find the container logs in var/log/messages and sos_commands/logs/journalctl_--no-pager_--boot, as expected. Thus I don't think, there is anything to do in particular. I'll update the doc field for that, but I don't see what else can be done. Thanks. Greg, does that work for you?
Thanks for checking on this Seb.! I am satisfied
LGTM, thanks Bara.