Skip to content

CASSANDRA-21693 Avoid failing concurrent node decommissions due to starting several reconfigurecms at the same time - #5186

Open
nivykani wants to merge 1 commit into
apache:trunkfrom
nivykani:node-decom
Open

nivykani wants to merge 1 commit into
apache:trunkfrom
nivykani:node-decom

Conversation

@nivykani

Copy link
Copy Markdown

Only one CMS reconfiguration can be active at a time, so operations that invoke maybeReconfigureCms (like decommissioning CMS nodes) will fail immediately if they run in parallel, instead of waiting and retrying.

This patch adds retries with exponential backoff and jitter up to a configurable timeout (cms_reconfiguration_wait_timeout, default 10m). The error on timeout is kept the same (IllegalStateException) and we also track the number of retries via cmsReconfigurationRetries.

The Cassandra Jira

# Conflicts:
#	CHANGES.txt
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant