Le 25/09/2014 07:49, Michael Biebl a écrit : > Am 25.09.2014 um 07:31 schrieb Raphaël HALIMI: >> Le 24/09/2014 17:30, Michael Biebl a écrit : >>> Am 24.09.2014 um 17:23 schrieb Raphaël HALIMI: > >>>> http://cgit.freedesktop.org/systemd/systemd/commit/?id=5e8b767df6e18444d5aff2987b5e5603361ed528 >>> >>> Could you verify that this patch does indeed fix the issue you're seeing? >> >> Sorry for not testing the patch before suggesting it. Unfortunately, it >> does not fix the problem. I tried to flood the journal with systemd-cat >> and a lot of messages were lost; but even with normal traffic (like when >> I stumbled on the bug in the first place), it still happens sometimes. > > Did you restart the socket and the syslog service (or reboot the system)?
I couldn't restart the *socket* (systemd stops it, but doesn't start it afterwards because the *service* is already active; maybe this fact deserves its own bug report). After that, restarting the service restarts both the service and the socket (only if the socket is stopped; if not, restarting the service doesn't restart the socket). With such a strange behavior, I had to check this explicitely by verifying the modification time of /run/systemd/journal/dev-log each time I stopped the socket and restarted systemd-journald. I didn't restart syslog, since I thought the problem comes from systemd-journald. Anyway, since this method didn't show any improvement with the fix applied, I rebooted the system to be sure; but unfortunately the results were the same. So yes, I did restart the remevant units and I can confirm that the fix adopted upstream doesn't really fix the problem. Regards, -- Raphaël HALIMI
signature.asc
Description: OpenPGP digital signature
