The wdqs indexer pipeline failed due to an existing pod.

The WMF maintains WMFKubernetesPodOperator as a sub-class of the KubernetesPodOperation. This sub-class will add a randomized suffix to the pod name, avoiding issues like this one.

While using the WMFKubernetesPodOperator will not solve the problem of the existing (garbage?) pod by the same name, it will at least avoid subsequent runs of the pipeline running into the same problem.

AC:

• The dump tasks in the indexer pipeline use KubernetesPodOperator with enabled random suffix support
• The cause of the duplicate pod has been determined and resolved. (we will keep an eye on the pods, but at this point there is no way to investigate, seeing as the previous offender has been removed in order to fix the pipeline)