Open julsemaan opened 7 years ago
Not sure that's something we can fix. I think systemd is to blame for that. Not that it ever makes any mistake, but still.
Unless I misunderstand the issue?
Our processes are the only ones that present that behavior.
Usually, I've seen systemd say that the process couldn't start when it fails. Maybe we need to implement a signal that says 'hey we've started properly' and prevent systemd from declaring the process is alive without that during startup
I'll try to dig into it a bit but feel free to check it out as well
@julsemaan will check if easy fix, it not, considered as low-priority and push to 7.1
@louismunro is right. No easy way I can see of checking that.
Moving to 7.1 for further investigation
sdnotify for apache (RHEL vs Debian)
@louismunro what's missing ?
I think haproxy is still missing, @louismunro can confirm
The following services don't support sd_notify yet:
packetfence-carbon-cache.service packetfence-carbon-relay.service packetfence-collectd.service packetfence-haproxy.service packetfence-keepalived.service packetfence-p0f.service packetfence-radsniff.service packetfence-redis-cache.service packetfence-redis_ntlm_cache.service packetfence-redis_queue.service packetfence-routes.service packetfence-statsd.service
We probably don't care much about most of the services above, but haproxy and keepalived would be nice. Unfortunately I think that would require patching them...
Can we remove the 7.2 milestone from this one ?
Make it PF 8.0 ?
I like your style !
Is there an update on this? Seeing it on Debian 9, Packetfence 10.2.0+20201007132.
When a systemd service fails to start (socket bind fail, compile error, etc) no error is shown and you actually have to look at the status to know whether it is alive or dead.
This is an uncommon behavior for systemd services, we should signal systemd a service is started only when it has actually started successfully.