[systemd-devel] Journald stops logging when lots of output

[Lennart Poettering]

On Fri, 27.09.13 09:47, Colin Guthrie (gmane at colin.guthr.ie) wrote:

> 
> 'Twas brillig, and Lennart Poettering at 26/09/13 19:35 did gyre and gimble:
> > On Thu, 26.09.13 15:46, Olav Vitters (olav at vitters.nl) wrote:
> > 
> >>
> >> On Wed, Sep 11, 2013 at 08:35:49PM +0200, Lennart Poettering wrote:
> >>> (Of course, journald should not exit under any such circumstances, but
> >>> to find that we first need to track down why it does that currently).
> >>
> >> Though it might hide problems, shouldn't journald be configured to
> >> automatically be restarted by systemd?
> > 
> > We do that, but not indefinitely. See
> > StartLimitInterval=/StartLimitBurst= in systemd.service(5).
> 
> Quick question, I presume logging is still lost for services stderr/out
> etc. when journald is restarted? Or has this been worked-around these days?

Still doesn't work. I do have an idea now however we could make this
work by using a single AF_UNIX/SOCK_DGRAM socket instead of individual
AF_UNIX/SOCK_STREAM sockets. For services we start it shouldn't matter
if stdout/stderr is SOCK_STREAM or SOCK_DGRAM really, and if we use
SOCK_DGRAM we even get per-client SCM_CREDENTIALS which would allow us
to properly attribute every single write to the right process which we
currently can't.

Lennart

-- 
Lennart Poettering - Red Hat, Inc.