filebeat is restarted on every highstate run

[kevinschmidt]

Since #5 , the filebeat.service is a cmd.run without unless which means it's run on every highstate run. As I'm running highstate every 5 minutes, filebeat restarts every 5 minutes.

I changed cmd.run to cmd.wait locally in my repo but that unfortunately means the service is only started on config file changes (because of watch_in) but not if it is not running e.g. because of a crash. I'm not sure what a good way to solve this would look like.

[gravyboat]

Let's bring @cybacolt into this discussion. I feel like this is a tough one to solve. Do you know if filebeat ever plans on creating an actual service? Cybacolt and I discussed this a bit before and that issue he linked is STILL open.

[cybacolt]

apologies for not noticing this earlier...

i initially built it using service, but due to some bugs in how filebeat daemon starts, i had to cmd.run with a ssh loopback is the only way it worked correctly on debian/ubuntu.

the unless: check, had other consequences - it wouldnt restart if already running, which is also not desirable if config has changed.

i agree it needs a way to satisfy both scenarios.

[kevinschmidt]

As they issue is still not fixed upstream, the only thing I can think of is to have two service entries: one that does cmd.run with unless and one that does cmd.wait triggered by a config file change. Not ideal but might work until upstream is fixed.

[akumria]

This appears to have been fixed in the upstream go-daemon (see: fiorix/go-daemon#7) and in filebeat v1.2.2 (see: https://www.elastic.co/guide/en/beats/libbeat/current/release-notes-1.2.2.html)

[blbradley]

@akumria I was just looking for that information. Thank you for the info.

[blbradley]

This is due to go-daemon being used in Debian init scripts when it is not required. Elastic may not update these to fix it.

[blbradley]

I've filed an issue upstream at elastic/beats#3053. We've determined that there is an issue on Debian but not what the issue actually is.