Open
Description
Hello
ydbops utility uses all nodes that was given by CMS (Cluster Management System) of YDB for rolling restart. For now in CMS users can set limits on the number of unavailable nodes in the database or storage. However, they cannot filter them by location and this can have a bad effect on query latency, especially when restarting databases when using the prefer-local-dc setting with YDB SDK inside user application.
I would like to propose a feature request that will allow the ydbop to limit the number of nodes that can be restarted within one location. This limit should work within the current window by --nodes-inflight
arg
Metadata
Assignees
Labels
No labels
Activity