making multi line work? Dead lock recursive locking

[Vince-Cercury]

I have some events on multi line

2019-03-18 06:06:48.859  INFO [manage-xxx-service,,,] [10.2.7.19] 1 --- [-15276-thread-1] o.a.k.clients.consumer.ConsumerConfig    : ConsumerConfig values:
        auto.commit.interval.ms = 5000
        auto.offset.reset = latest
        bootstrap.servers = [my-kafka-service:9092]
        check.crcs = true

So I added this line:

  manage-xxx-service:
    <<: *glog
    from:
      pod: my-xxx-service
    multiline:
      firstline: /^\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2}/
    sourcetype: kube:my-xxx-service

Which produces this config:

      <filter tail.containers.var.log.containers.my-xxx-service*my-xxx-service*.log>
        @type concat
        key log
        timeout_label @SPLUNK
        stream_identity_key stream
        multiline_start_regexp /^\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2}/
        flush_interval 5s
      </filter>

I'm getting an error "deadlock; recursive locking"

2019-03-18 05:13:54 +0000 [warn]: #0 dump an error event: error_class=ThreadError error="deadlock; recursive locking" location="/usr/local/bundle/gems/fluent-plugin-concat-2.3.0/lib/fluent/plugin/filter_concat.rb:144:in `synchronize'" tag="tail.containers.var.log.containers.my-xxx-service-85855985fc-pgl6g_yyy_my-incident-service-0ee1814dcd3596c96e0bf6c0a2e65a9437cf1b282a95daf41fbd6e8933df1f8f.log" time=

What am I doing wrong?

[Vince-Cercury]

Note that despite the error, it works. But I don't think I should proceed until error is gone

[matthewmodestino]

Thanks for reporting this Vince, and appreciate you logging the issue on the concat repo.

fluent-plugins-nursery/fluent-plugin-concat#69

I will assist with following up with the concat devs, in the meantime, I will validate that flipping back to the 1.0.1 logging images doesn't show this error.

The 1.1.0 images bumped the fluentd and concat plugin versions, so it may just be that they don't play nice yet with fluent 1.4.

[Vince-Cercury]

I can confirm 1.0.1 removed the errors

[matthewmodestino]

I have updated the issue on the concat repo with reproducible steps for docker.

[matthewmodestino]

fluent-plugins-nursery/fluent-plugin-concat#69 (comment)

[chaitanyaphalak]

Fixed in 1.2.0

[mebuzz]

I am still having this issue on SplunkConnect for kuberentes(i am on AWS EKS). Do I update the fluentd image to the updated one -1.2.0?

[matthewmodestino]

yes, update to fluentd-hec image 1.1.1 OR see link to concat issues where there is a label workaround. looks like root cause is due to events being remitted to @splunk label...

[mebuzz]

Thank you for the super quick response @matthewmodestino .
I checked the fluentd-hec-image version. It is updated to 1.1.1. My values.yaml contains the following lines:
image:
name: splunk/fluentd-hec:1.1.1

Should I just update the version here to "1.2.0"??

[matthewmodestino]

no, there is no 1.2

https://hub.docker.com/r/splunk/fluentd-hec/tags

Can you share the log you are seeing and the config you are running?

[mebuzz]

You mean the pod logs ?? And the config of configMap or the daemonset?

[matthewmodestino]

yes, I would pick a logging pod and dump the logs. At the head of the log we render the configuration the pod is running, then grab one of the errors you are seeing.

[mebuzz]

I am not seeing any errors in the logs.

[matthewmodestino]

ok...so let's back up then. Are you setting multiline rules in your configmap? If so, are you seeing the deadlock error?

[mebuzz]

Yes, we have put in this config in the configmap for multiline:
<filter tail.containers.var.log.containers.*.log>
@type concat
key log
timeout_label @splunk
stream_identity_key stream
multiline_start_regexp /^\d{4}-\d{2}-\d{2}\s\d{2}:\d{2}:\d{2}.\d{3}\d/
flush_interval 5s

[matthewmodestino]

and are you seeing errors? also do all your containers actually match that line breaker?

[mebuzz]

I have pushed the settings via configmap, so all containers should be having these settings.