Ubuntu
systemd package

  1. Bug #1433320
  2. Comment #10

Comment 10 for bug 1433320

Revision history for this message
Martin Pitt (pitti) wrote : Re: systemd-journald, udev, logind get killed by 1 min watchdog timeout

There is nothing specific to epoll_wait(), that's just the function that these daemons wait in for practially all the time. The SIGABRT is due to systemd killing/restarting the processes as they don't respond withing 1 minute (watchdog). In Fila's case the kernel reports hung tasks (in a page fault, i. e. slow/broken swap?), but in Till's reports there is no other apparent error message.

This is not just limited to QEMU, e. g. duplicate bug 1438805 is a real machine with a hardware watchdog. So we still only merely know the symptoms -- something stalls these QEMU or real laptop processes, but we don't know yet how. Does anyone have some observations/recipes how to reproduce this? I. e. did you start a lot of QEMU instances in parallel or something else what could make these slow? Maybe a running SSD TRIM in the background?