[
https://issues.apache.org/jira/browse/HADOOP-16349?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Steve Loughran updated HADOOP-16349:
------------------------------------
Description:
If you delete the version marker from a S3Guard table, it appears to hang for 5
minutes.
Only if you restart and turn logging to debug do you see that
{{DynamoDBMetadataStore.getVersionMarkerItem()}} is sleeping and retrying.
# log at warn
# add entry to troubleshooting doc on the topic
The cause of the failure can be any of
* table being inited elsewhere: expectation, fast recovery
* it's not a S3Guard table: it won't recover
* it's a S3Guard table without a version marker: it won't recover.
+ consider having a shorter retry lifespan, though if it adds a new config
point I'm a bit reluctant. For s3guard bucket-info it would make sense to
change the policy to be aggressively short lived
was:
If you delete the version marker from a S3Guard table, it appears to hang for 5
minutes.
Only if you restart and turn logging to debug do you see that
{{DynamoDBMetadataStore.getVersionMarkerItem()}} is sleeping and retrying.
# log at warn
# add entry to troubleshooting doc on the topic
The cause of the failure can be any of
* table being inited elsewhere: expectation, fast recovery
* it's not a S3Guard table: it won't recover
* it's a S3Guard table without a version marker: it won't recover.
> DynamoDBMetadataStore.getVersionMarkerItem() to log at info/warn on retry
> -------------------------------------------------------------------------
>
> Key: HADOOP-16349
> URL: https://issues.apache.org/jira/browse/HADOOP-16349
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
> Affects Versions: 3.2.0
> Reporter: Steve Loughran
> Assignee: Steve Loughran
> Priority: Major
>
> If you delete the version marker from a S3Guard table, it appears to hang for
> 5 minutes.
> Only if you restart and turn logging to debug do you see that
> {{DynamoDBMetadataStore.getVersionMarkerItem()}} is sleeping and retrying.
> # log at warn
> # add entry to troubleshooting doc on the topic
> The cause of the failure can be any of
> * table being inited elsewhere: expectation, fast recovery
> * it's not a S3Guard table: it won't recover
> * it's a S3Guard table without a version marker: it won't recover.
> + consider having a shorter retry lifespan, though if it adds a new config
> point I'm a bit reluctant. For s3guard bucket-info it would make sense to
> change the policy to be aggressively short lived
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
