[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Support for Kubernetes metadata when using docker --log-driver=journald?



Per cluster.

On Fri, May 20, 2016 at 12:46 PM, Rich Megginson <rmeggins redhat com> wrote:
> On 05/20/2016 10:39 AM, Clayton Coleman wrote:
>>
>> Actually, there may be an issue for this already.
>> https://github.com/kubernetes/kubernetes/issues/15478
>
>
> I created a new issue: https://github.com/kubernetes/kubernetes/issues/25975
>
>
>>
>> On Fri, May 20, 2016 at 12:38 PM, Clayton Coleman <ccoleman redhat com>
>> wrote:
>>>
>>> We could support setting that on the Kubelet when we launch the
>>> containers.  I think it's reasonable - if you open a Kube issue Derek
>>> and I can chime in.
>>>
>>> On Fri, May 20, 2016 at 12:12 PM, Rich Megginson <rmeggins redhat com>
>>> wrote:
>>>>
>>>> When using OpenShift with the standard docker log driver, there are
>>>> container logs in the following format:
>>>>
>>>>
>>>> /var/log/containers/$name_of_pod-$name_of_project-$name_of_container-$containeruuid.log
>>>>
>>>> The fluentd log collector uses $name_of_pod and $name_of_project to look
>>>> up
>>>> in k8s additional metadata for the pod and project (such as the uuids).
>>>>
>>>> When using docker daemon --log-driver=journald, these files no longer
>>>> exist,
>>>> and the journal entries are missing $name_of_pod and $name_of_project in
>>>> the
>>>> metadata:
>>>>
>>>>
>>>> __CURSOR=s=7c08f275b8a545a9afb1770ef779ebb8;i=3148;b=caa653694cfa4a6ea9e4213e6cf4065b;m=13700ce9;t=53347368f6904;x=770aebf7b30ab7b3
>>>> __REALTIME_TIMESTAMP=1463755834616068
>>>> ...
>>>> _COMM=docker-current
>>>> _EXE=/usr/bin/docker-current
>>>> _CMDLINE=/usr/bin/docker-current daemon
>>>> --authorization-plugin=rhel-push-plugin --selinux-enabled
>>>> --log-driver=journald
>>>> CONTAINER_ID=a160de6fe8fa
>>>>
>>>> CONTAINER_ID_FULL=a160de6fe8facb49103c45f2bd14854b455b42153b777634ea58fb8a95124a40
>>>> CONTAINER_NAME=silly_bhaskara
>>>>
>>>> When running containers via docker directly, you can use -e and -l to
>>>> pass
>>>> additional metadata through to the journal output:
>>>>    docker run -e FOOT=BALL -l LA=SF --rm --name toasted \
>>>>        --log-driver=journald --log-opt env=FOOT --log-opt labels=LA \
>>>>        fedora /bin/pwd
>>>>
>>>>   MESSAGE=/
>>>>   CONTAINER_NAME=toasted
>>>>   LA=SF
>>>>   FOOT=BALL
>>>>   CONTAINER_ID=37772f57b118
>>>>
>>>>
>>>> CONTAINER_ID_FULL=37772f57b1189306fed624675cd33412461a6f04ff0f4d7a4ea5d1d41f637641
>>>>
>>>> Is there some way to make k8s and/or OpenShift add fields like
>>>> K8S_POD_NAME=$name_of_pod and
>>>> K8S_NAMESPACE_NAME=$name_of_project to the journal entries?
>>>> This will also be needed by
>>>> https://github.com/fabric8io/fluent-plugin-kubernetes_metadata_filter
>>>> which
>>>> assumes the entries are tagged in a similar format to the filename
>>>> above.
>>>> Of course if k8s/openshift can be configured to automatically add that
>>>> additional metadata to the journal entries then the fluentd k8s metadata
>>>> plugin might not be needed.
>>>>
>>>>
>>>> _______________________________________________
>>>> dev mailing list
>>>> dev lists openshift redhat com
>>>> http://lists.openshift.redhat.com/openshiftmm/listinfo/dev
>>>>
>
> _______________________________________________
> dev mailing list
> dev lists openshift redhat com
> http://lists.openshift.redhat.com/openshiftmm/listinfo/dev


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]