[
https://issues.apache.org/jira/browse/HDDS-5220?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17345898#comment-17345898
]
Lokesh Jain commented on HDDS-5220:
-----------------------------------
[~cchenax] Thanks for finding the root cause!
I think changing the deletion order might not help. Let's consider a scenario.
We have a deletion thread and scanner thread.
1. Scanner fetches the block metadata.
2. Deletion thread deletes both metadata and data. (Any order)
3. Comparing checksum for data in scanner would fail now.
I think the we might need to handle block deletion case in scanner.
> when datanode delete the data,first delete metadata,then delete chunk files
> ---------------------------------------------------------------------------
>
> Key: HDDS-5220
> URL: https://issues.apache.org/jira/browse/HDDS-5220
> Project: Apache Ozone
> Issue Type: Improvement
> Reporter: cchenaxchen
> Assignee: cchenaxchen
> Priority: Major
> Labels: pull-request-available
>
> when use datanode scanner scan the container，the checksum bases on Rocksdb
> metadata to find the chunk file on disk,but when datanode delete
> data，datanode should delete metadata first，the delete chunk file，in order to
> avoid when chunk file was deleted，but the metadata can not be modified，the
> datanode scanner scan the datanode will mark this container unhealthy.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
