[
https://issues.apache.org/jira/browse/IGNITE-20287?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mirza Aliev updated IGNITE-20287:
---------------------------------
Description:
h3. *Motivation*
We need to clean up resources for destroyed distribution zones from the
catalog. It is possible that while a zone is removed, some actions that must be
done on a zone deletion could be interrupted by restart. On recovery, we must
detect such zone's deletion an must clean up the resources. Currently we store
some zone's state in Meta Storage, so this resources must be cleaned up.
h3. *Definition of done*
There is no
h3. *Implementation notes*
For zones that are not presented in the catalog, but presented in the MS, just
remove all data nodes related keys. All these changes must be done using meta
storage condition which we us when we call
{{DistributionZoneManager#removeTriggerKeysAndDataNodes}} on a zone drop.
was:
h3. *Motivation*
We need to clean up resources for destroyed distribution zones from the
catalog. It is possible that while a zone is removed, some actions that must be
done on a zone deletion could be interrupted by restart. On recovery, we must
detect such zone's deletion an must clean up the resources. Currently we store
some zone's state in Meta Storage, so this resources must be cleaned up.
h3. *Definition of done*
h3. *Implementation notes*
For zones that are not presented in the catalog, but presented in the MS, just
remove all data nodes related keys. All these changes must be done using meta
storage condition which we us when we call
{{DistributionZoneManager#removeTriggerKeysAndDataNodes}} on a zone drop.
> Clean up removed resources for destroyed zones in catalog
> ---------------------------------------------------------
>
> Key: IGNITE-20287
> URL: https://issues.apache.org/jira/browse/IGNITE-20287
> Project: Ignite
> Issue Type: Improvement
> Reporter: Kirill Tkalenko
> Priority: Major
> Labels: ignite-3
>
> h3. *Motivation*
> We need to clean up resources for destroyed distribution zones from the
> catalog. It is possible that while a zone is removed, some actions that must
> be done on a zone deletion could be interrupted by restart. On recovery, we
> must detect such zone's deletion an must clean up the resources. Currently we
> store some zone's state in Meta Storage, so this resources must be cleaned up.
> h3. *Definition of done*
> There is no
> h3. *Implementation notes*
> For zones that are not presented in the catalog, but presented in the MS,
> just remove all data nodes related keys. All these changes must be done using
> meta storage condition which we us when we call
> {{DistributionZoneManager#removeTriggerKeysAndDataNodes}} on a zone drop.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
