On 24.02.2015 13:14, Rich Freeman wrote: > I suspect this is trivial - it looks like something like this would work: > http://.../entries?_SYSTEMD_UNIT=postfix.service
Yes, correct, as I thought this is the easy part. Works: http://mythtv.local:19531/entries?_SYSTEMD_UNIT=postfix.service (using my mythbackend as test target). > (note, you might need to tweak that - I haven't used the http gateway > personally and am going from the manpage) > >> AND only the lines relevant for the domain of the specific customer by >> doing this? > > I think you're going to be stuck here unless they come from different > machines or something like that. Obviously you can pipe the output > through grep but journald will only pre-filter the output using > journal fields, like facility, priority, etc. syslog only provided a > few fields for clients to specify, and this is probably because in the > end the data just got dumped to a text file so that it wasn't > searchable by field anyway. It would be nice if they extended the > syslog protocol for systemd and made it possible for clients to > specify additional fields, but obviously the client would need to > support this (likely sending logs over dbus or such). > > The http gateway seems like it is intended more as a transport > mechanism with some usability for ad-hoc human viewing. It isn't a > full-fledged log analysis tool. The fact that journald can output in > JSON with uuids for each entry should make it far easier to parse its > logs with an analysis tool, but I think all those vendors are playing > catch-up. I suspect they'll support it fairly soon once they see > everybody using it. From a machine parsing standpoint the fielded > binary format makes a lot more sense. Maybe I could set up some other web-app that (a) looks at the link pointing to the postfix.service-logs and (b) filters them? I could post to the systemd-devel-ml ... btw ;-) Stefan
