How to trigger cleanup tasks sequentially instead of in parallel on 1.12

Kubernetes Operator
, , ,
1

Hi, I’m using Scylla Operator 1.12. After scaling, cleanup jobs seem to run simultaneously on all nodes, which puts significant pressure on the cluster. Wonder is there a way to run these jobs sequentially instead of in parallel?

2

@Yehuda_Lebi take a look…

3

The Scylla Operator automatically kicks off cleanup jobs after a ScyllaCluster scale event. It creates per-node Jobs in parallel, there’s no user knob to serialize them today.

1 Like
4

Thanks for the reply. Not sure but I feel such an option might be necessary, as the cluster could be under heavy load overall if the dataset is large.

5

Thanks for the feedback. I’ve opened Add option to control cleanup parallelism · Issue #3015 · scylladb/scylla-operator · GitHub as a tracker feature request.

2 Likes