Description of problem: At least since the latest lvm2 update (lvm2-2.02.166-1.el7_3.4.x86_64) Yum spends a long time (~90 seconds) in the clean up phase of lvm2; pstree(1) shows a postscript is executing "systemctl try-restart lvm2-lvmetad.service". This has been observed on some machines, and even on them sporadically, doing "systemctl try-restart lvm2-lvmetad.service" afterwards may get stuck once but then work ok later on. Apr 19 10:00:01 localhost systemd: Stopping LVM2 metadata daemon... Apr 19 10:00:01 localhost lvmetad[569]: Failed to accept connection errno 11. Apr 19 10:01:31 localhost systemd: lvm2-lvmetad.service stop-sigterm timed out. Killing. Apr 19 10:01:31 localhost systemd: lvm2-lvmetad.service: main process exited, code=killed, status=9/KILL Apr 19 10:01:31 localhost systemd: Unit lvm2-lvmetad.service entered failed state. Apr 19 10:01:31 localhost systemd: lvm2-lvmetad.service failed. Apr 19 10:01:31 localhost systemd: Started LVM2 metadata daemon. Apr 19 10:01:31 localhost systemd: Starting LVM2 metadata daemon... Apr 19 10:01:31 localhost systemd: Reloading. Please try to make the "systemctl try-restart lvm2-lvmetad.service" operation faster in all cases. Version-Release number of selected component (if applicable): lvm2-2.02.166-1.el7_3.4.x86_64
It is reproducible. This seems the problem: > Apr 19 10:00:01 localhost lvmetad[569]: Failed to accept connection errno 11.
Similar issue was dealt with here https://sourceware.org/git/?p=lvm2.git;a=commitdiff;h=b12961e7ebd6fb29d760daafd3c16e4ba3e54e80 An lvmetad core or lvmetad gdb backtrace would probably help.
I tried to reproduce this by downgrading lvm2/device-mapper packages on few machines where I saw this and then upgrading again but the issue did not occur. Currently I am unable to provide more information. Thanks.
(In reply to Marian Csontos from comment #2) > It is reproducible. This seems the problem: > > > Apr 19 10:00:01 localhost lvmetad[569]: Failed to accept connection errno 11. Marian, if you can reproduce, can you provide the info dave is looking for?
"Reproducible" is not the right word. "Common" would be better. Second attempt to try-restart was executed immediately. If seen again I will try to get more details.
No data collected in two years. Closing this...