[Bug 1696970] Re: softlockup DoS causes systemd-journald.service to abort with SIGABORT

[Balint Reczey]

@colin-king In this case the journal was not broken. Do you have many
examples of really broken journals?

If so, do you have any solution in mind that would work better in
scenarios you care about but don't cause regressions?

Proper infrastructure planning with (CPU) quotas in place where it
matters seems to be a good practice instead of running highly loaded
servers.

The 3 min default timer for systemd-journald seems appropriate or even
generous.

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to systemd in Ubuntu.
https://bugs.launchpad.net/bugs/1696970

Title:
  softlockup DoS causes systemd-journald.service to abort with SIGABORT

Status in systemd package in Ubuntu:
  Opinion
Status in systemd source package in Artful:
  Opinion

Bug description:
  I was running the new stress-ng softlockup stressor and observed that
  systemd-journald gets killed with an abort and this corrupts the
  systemd journal.

  How to reproduce:

  git clone git://kernel.ubuntu.com/cking/stress-ng
  cd stress-ng
  make clean; make

  sudo ./stress-ng --softlockup 0 -t 360 -v

  ..and wait for 360 seconds.  dmesg shows the following, 100%
  reproduceable:

  
  [  875.310331] systemd[1]: systemd-timesyncd.service: Watchdog timeout (limit 3min)!
  [  875.310740] systemd[1]: systemd-timesyncd.service: Killing process 574 (systemd-timesyn) with signal SIGABRT.
  [  875.327289] systemd[1]: systemd-timesyncd.service: Main process exited, code=killed, status=6/ABRT
  [  875.327666] systemd[1]: systemd-timesyncd.service: Unit entered failed state.
  [  875.327686] systemd[1]: systemd-timesyncd.service: Failed with result 'watchdog'.
  [  875.327917] systemd[1]: systemd-timesyncd.service: Service has no hold-off time, scheduling restart.
  [  875.327954] systemd[1]: Stopped Network Time Synchronization.
  [  875.328845] systemd[1]: Starting Network Time Synchronization...
  [  875.525071] systemd[1]: Started Network Time Synchronization.
  [  875.539619] systemd[1]: systemd-journald.service: Main process exited, code=dumped, status=6/ABRT
  [  875.544257] systemd-journald[5214]: File /run/log/journal/440e485e550040e3b93b66b2faae8525/system.journal corrupted or uncleanly shut down, renaming and replacing.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/1696970/+subscriptions