Open hamsterready opened 7 years ago
Thanks! Fixed, let me know how it works out.
When monitored containers are restarted/recreated then I see nice:
ubuntu@backend-03:~$ docker logs -f e817f9b3ce90 2>&1 | grep Processing
Processing rta_metrics-consumer.d5fvj98is8xookf2l8534isse.svg2txg6i3jvk35aa5ndsce6o ...
Processing rta_metrics-api.d5fvj98is8xookf2l8534isse.7u0d49rre4agmtxytt7jrnfej ...
Processing rta_metrics-consumer.d5fvj98is8xookf2l8534isse.3ipv9bi3mrnbmnzp6nicxfg8e ...
Processing rta_metrics-consumer.d5fvj98is8xookf2l8534isse.k19hs64jk4xu7nu70oowqph9f ...
Processing rta_metrics-consumer.d5fvj98is8xookf2l8534isse.lnphgm20jknzrvk8f84xl9wee ...
Processing rta_metrics-consumer.d5fvj98is8xookf2l8534isse.xle9xx76md5unczpt08vbx4f5 ...
instead of
Err State for should have been dropped, but couldn't as state is not finished.
Cheers,
hi @willfarrell
I think I was a bit too early with confirming all is good.
The tail -f
on fifo works in a bit different way then I was expected https://stackoverflow.com/questions/20409071/tail-f-on-a-named-pipe-behaves-strange and as such when tail is being used one will lose the data.
On the other hand when cat
is being used the filebeat will crash/hang on container restart and will stop delivering logs to logstash or whatever is configured.
In essence the filebeat becomes useless in docker environment - lovely.
So ... we have to choose the lesser of two evils.
Man, to be honest I decided to go with logspout... and then was hit by this issue https://github.com/gliderlabs/logspout/issues/330 bad luck
In the wrapper script:
https://github.com/willfarrell/docker-filebeat/blob/e1b40b46dc9edca80e37527f26bbf2c8e61917fe/5-stdin/docker-entrypoint.sh#L29
I would use use
tail -f
instead ofcat
.Otherwise on container restart I was experiencing issue similar to this one: https://github.com/elastic/beats/issues/3315
with error message;