[
https://issues.apache.org/jira/browse/CASSANDRA-20267?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Marcus Eriksson updated CASSANDRA-20267:
----------------------------------------
Test and Documentation Plan: ci run
Status: Patch Available (was: Open)
> Reduce heap pressure when initializing the CMS
> ----------------------------------------------
>
> Key: CASSANDRA-20267
> URL: https://issues.apache.org/jira/browse/CASSANDRA-20267
> Project: Apache Cassandra
> Issue Type: Bug
> Components: Transactional Cluster Metadata
> Reporter: Marcus Eriksson
> Assignee: Marcus Eriksson
> Priority: Normal
> Fix For: 5.x
>
> Attachments: ci_summary.html
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> When running {{nodetool cms initialize}} we currently request the full
> gossip-initialized cluster metadata from all replicas in the cluster and
> compare them. This doesn't work in large clusters.
> Instead the CMS initiator should send out a reduced version of its cluster
> metadata to all replicas (excluding the placements because they can be
> calculated based on tokenMap+schema+directory). The replicas then reply
> "yes/no" if their local metadata matches.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
