Closed dannyk81 closed 7 years ago
I am not sure what the best way to handle this would be ....
How do other plugins handle this sort of fatal error?
There are two real options to handle this
I think it should log a warning and retry to access, similar approach that the tail plugin uses (I believe):
2017-01-10 03:38:11 +0000 [warn]: /var/log/xyz.log unreadable. It is excluded and would be examined next time.
I could go for that, doing the same thing as the tail plugin makes a lot of sense.
Did you wan't to have a bash at this @dannyk81 ? Otherwise it will go onto my list :)
I could try :) but no much experience with Ruby I'm afraid...
By the way, slightly unrelated... the new 0.1.0 requires fluentd 0.14.11 ?
By the way, slightly unrelated... the new 0.1.0 requires fluentd 0.14.11 ?
Yes. If you are still using Fluentd 0.12.x, please use fluent-plugin-systemd 0.0.5.
see: https://github.com/reevoo/fluent-plugin-systemd/pull/25/files#diff-04c6e90faac2675aa89e2176d2eec7d8R11 see: https://github.com/reevoo/fluent-plugin-systemd/blob/fa5f1ebd595ee56395e5c7a270bb1113e62e0d05/fluent-plugin-systemd.gemspec#L26
Just my thought, Fluentd and its plugins should detect any suspicious and erroneous settings in #configure
instead of after #start
. Because it helps users to notice errors and warnings before launching completely.
By the way, slightly unrelated... the new 0.1.0 requires fluentd 0.14.11 ?
I have just cut a 0.0.x branch for maintenance of a v0.12 compatible version hopefully we can continue to backport stuff there at least until td-agent is based on v0.14
I have just cut a 0.0.x branch for maintenance of a v0.12 compatible version hopefully we can continue to backport stuff there at least until td-agent is based on v0.14
Thanks! that would be a awesome.
@errm I've found that this line should be ~> 0.12.0
in 0.0.x branch.
Because rubygems interprets it as >= 0.12.0 && < 1.0
.
The right way to specify locking v0.12.x version is ~> 0.12.0
.
^^ I think you are right, but it shouldn't cause too much harm since 0.0.x should still work ok on v0.14
Awesome!!! Thanks for this :+1:
Using
fluent-plugin-systemd
version0.0.5
onfluentd-0.12.31
(Dockerized in K8s environment)When starting fluentd with this plugin, in case the journal path is not available on the server it causes fluentd to crash.
Any way to handle it more gracefully ?