[
https://issues.apache.org/jira/browse/CASSANDRA-21129?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18053798#comment-18053798
]
Abhijeet Dubey edited comment on CASSANDRA-21129 at 1/23/26 6:35 AM:
---------------------------------------------------------------------
Scenario: "Node was working, now it won't start"
With SSTableMetadataDump (this change)
# Point directly at the data directory
sstablemetadatadump dump --data-dir /var/lib/cassandra/data --text --all
What happens:
1. Finds SSTables in /var/lib/cassandra/data/system/local_metadata_log-/ and
metadata_snapshots-/
2. Imports them into an in-memory ColumnFamilyStore (no full node startup)
3. Reconstructs ClusterMetadata from the actual on-disk data
4. Displays epochs, schema, directory, tokens, etc.
So, You see the exact state that was persisted to disk before the crash.
---------------------------------------------------------------------------
With PR #4149 (CMSOfflineTool)
# Requires a dump file that was created BEFORE the node died
cmsofflinetool describe -f /path/to/previous-dump.dump
What happens:
1. Looks for a pre-existing dump file
2. If no dump exists -> Tool cannot help
3. If dump exists -> Shows metadata from when the dump was taken (potentially
stale)
was (Author: JIRAUSER306221):
Scenario: "Node was working, now it won't start"
With SSTableMetadataDump (this change)
# Point directly at the data directory
sstablemetadatadump dump --data-dir /var/lib/cassandra/data --text --all
What happens:
1. Finds SSTables in /var/lib/cassandra/data/system/local_metadata_log-*/ and
metadata_snapshots-*/
2. Imports them into an in-memory ColumnFamilyStore (no full node startup)
3. Reconstructs ClusterMetadata from the actual on-disk data
4. Displays epochs, schema, directory, tokens, etc.
So, You see the exact state that was persisted to disk before the crash.
---------------------------------------------------------------------------
With PR #4149 (CMSOfflineTool)
# Requires a dump file that was created BEFORE the node died
cmsofflinetool describe -f /path/to/previous-dump.dump
What happens:
1. Looks for a pre-existing dump file
2. If no dump exists -> Tool cannot help
3. If dump exists -> Shows metadata from when the dump was taken (potentially
stale)
> Tool to offline-dump the cluster metadata from an instance
> ----------------------------------------------------------
>
> Key: CASSANDRA-21129
> URL: https://issues.apache.org/jira/browse/CASSANDRA-21129
> Project: Apache Cassandra
> Issue Type: New Feature
> Components: Transactional Cluster Metadata
> Reporter: Abhijeet Dubey
> Assignee: Abhijeet Dubey
> Priority: Normal
> Attachments: text-output.txt, to-string-dump-output.txt
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> When a Cassandra node fails to start due to Transactional Cluster Metadata
> (TCM/CEP-21) corruption or issues, operators need a way to inspect the
> cluster metadata state offline without starting the node. The existing tools
> (nodetool, cqlsh) require a running node, leaving operators blind when
> debugging startup failures.
>
> When a node fails to start due to TCM corruption or inconsistencies,
> operators have no way to inspect the metadata state without a running node.
> This tool fills that gap by reading directly from SSTables.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
