I would rollback to 1.13.5 for now then

On Tue, Dec 11, 2018, 00:49 Simon Young <[email protected]>
wrote:

> Hi Carlos,
>
> Thanks for the suggestion. I have tried changing the "Max connections to
> Kubernetes API" and this does have an effect, but not in a way I'd
> expect... I've found that if I change that value in any way at all (either
> increase or decrease), then the next build always passes, and all
> subsequent builds fail.
>
> So for example, I tried raising the value to 200 via the Jenkins UI. I hit
> 'Apply' and re-ran the job and it passed, but it failed the next time it
> ran, so I raised the value to 400. Again, the next job passed, but any
> after that failed. At this point, I suspected that changing the value in
> any way might have this effect, so I reduced the value back to 32. Same
> thing happened - pass, then subsequent fails. I reduced the value all the
> way down to 1, and observed the same symptoms.
>
> I tried changing other plugin configuration values such as "Connection
> Timeout" and "Container Cap", but this did not have the same effect.
>
> It's also worth noting if I don't run the job for about an hour, the next
> build passes and subsequent builds fail. Same happens if I restart Jenkins
> (but not if I just reload the config from disk).
>
> So it seems like some internal counter on the Jenkins master is reaching a
> limit, and the counter is reset when the config changes or after a period
> of activity. Neither master nor slave appear to be running out of physical
> resources, and this is the only job we're running on the master.
>
> Have you any idea what could be going on here?
>
> Simon.
>
>
>
> On Monday, December 10, 2018 at 4:10:27 PM UTC, Carlos Sanchez wrote:
>>
>> You may be hitting the limit of concurrent connections to k8s api, see
>> https://github.com/jenkinsci/kubernetes-plugin/blob/master/CHANGELOG.md#1136
>>
>> On Mon, Dec 10, 2018 at 8:01 AM Simon Young <
>> [email protected]> wrote:
>>
> Hi,
>>>
>>> We are trying to use the Kubernetes Plugin to run tests in a different
>>> container - but *almost* every time we try and execute a command in the
>>> second container, the build fails immediately. I say "almost" because the
>>> build always succeeds after restarting the Jenkins master (or possibly
>>> after it's been idle for a long time).
>>>
>>> Here's the Jenkinsfile we're using to test:
>>>
>>> def label = "k8s-test-${UUID.randomUUID().toString()}"
>>>
>>> podTemplate(
>>>     cloud: 'kubernetes-test.k8s.local',
>>>     namespace: 'mynamespace',
>>>     label: label,
>>>     yaml: """
>>> apiVersion: v1
>>> kind: Pod
>>> spec:
>>>   containers:
>>>   - name: busybox
>>>     image: busybox
>>>     command: ['cat']
>>>     tty: true
>>> """
>>>     ) {
>>>
>>>     node (label) {
>>>         stage ('Get Agent Info') {
>>>             sh "java -version"
>>>         }
>>>         stage ('Run tests') {
>>>             container('busybox') {
>>>                 sh "echo foo"
>>>                 sh "netstat -tln"
>>>                 sh "df -h"
>>>             }
>>>         }
>>>         stage ('Grab Logs') {
>>>             containerLog('jnlp')
>>>         }
>>>     }
>>> }
>>>
>>> This fails at the "echo foo" step. The error reported by the pipeline is:
>>>
>>> Task okhttp3.RealCall$AsyncCall@b5277fa rejected from
>>> java.util.concurrent.ThreadPoolExecutor@6c383ec7[Terminated, pool size
>>> = 0, active threads = 0, queued tasks = 0, completed tasks = 3]
>>>
>>> What we know so far:
>>>
>>> * Both containers are successfully deployed to the same pod, as expected.
>>> * If we don't try and run the shell commands on the 'busybox' container,
>>> the build completes successfully.
>>> * If Jenkins is restarted, the build passes. But subsequent builds fail.
>>> * The Agent logs contain similar exceptions whether the builds pass or
>>> fail, so it's hard to pinpoint a cause.
>>>
>>> The ThreadPool exception may imply that *something* has run out of
>>> Executors, but it's not clear what, or how to rectify the situation.
>>>
>>> Has anyone seen anything like this? Any ideas how to get to the bottom
>>> of it? As per the plugin's README, I've created Jenkins log recorders for
>>> org.csanchez.jenkins.plugins.kubernetes and okhttp3, but there are no
>>> obvious problems reported in the logs.
>>>
>>> Software Versions:
>>>
>>> Jenkins: v2.150.1
>>> Kubernetes Plugin: 1.13.7
>>> Kubernetes: 1.10.11 (same behaviour observed on 1.8.7)
>>>
>>> All suggestions appreciated!
>>>
>>> Thanks,
>>>
>>> Simon.
>>>
>>> --
>>> You received this message because you are subscribed to the Google
>>> Groups "Jenkins Users" group.
>>>
>> To unsubscribe from this group and stop receiving emails from it, send an
>>> email to [email protected].
>>> To view this discussion on the web visit
>>> https://groups.google.com/d/msgid/jenkinsci-users/141b526e-f00b-434d-9f43-ec97fdb23df7%40googlegroups.com
>>> <https://groups.google.com/d/msgid/jenkinsci-users/141b526e-f00b-434d-9f43-ec97fdb23df7%40googlegroups.com?utm_medium=email&utm_source=footer>
>>> .
>>> For more options, visit https://groups.google.com/d/optout.
>>>
>> --
> You received this message because you are subscribed to the Google Groups
> "Jenkins Users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/jenkinsci-users/29389695-c3c2-415b-b945-e1b4a9099fbe%40googlegroups.com
> <https://groups.google.com/d/msgid/jenkinsci-users/29389695-c3c2-415b-b945-e1b4a9099fbe%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
> For more options, visit https://groups.google.com/d/optout.
>

-- 
You received this message because you are subscribed to the Google Groups 
"Jenkins Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/jenkinsci-users/CALHFn6OXSoX0vzZxKhoYTNA0JRcAi-BzOsZvPTVO3gksBrGCxA%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to