[
https://issues.apache.org/jira/browse/HDDS-12595?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17939033#comment-17939033
]
Ethan Rose commented on HDDS-12595:
-----------------------------------
Adding some ideas for the implementation:
* Users would need to authenticate as the SCM admin for this command to work
* We can get the SCM container state from the
[getContainer|https://github.com/apache/ozone/blob/2b48e8c6ec1739d541d5c02183ad1a91d9f7a308/hadoop-hdds/common/src/main/java/org/apache/hadoop/hdds/scm/client/ScmClient.java#L71]
API in SCM.
* We could use the
[getContainerReplicas|https://github.com/apache/ozone/blob/2b48e8c6ec1739d541d5c02183ad1a91d9f7a308/hadoop-hdds/common/src/main/java/org/apache/hadoop/hdds/scm/client/ScmClient.java#L89]
SCM API for replica info, but it is probably better to get this information
from the datanodes since that is how the other replica verification checks work.
* To get the information from the datanodes we can use the [readContainer
API|https://github.com/apache/ozone/blob/b91292576e6a13c3c0149aca06847397b6ccc5a6/hadoop-hdds/common/src/main/java/org/apache/hadoop/hdds/scm/storage/ContainerProtocolCalls.java#L667]
on the datanodes.
** We can use
[ContainerOperationClient#readContainer|https://github.com/apache/ozone/blob/2b48e8c6ec1739d541d5c02183ad1a91d9f7a308/hadoop-hdds/tools/src/main/java/org/apache/hadoop/hdds/scm/cli/ContainerOperationClient.java#L374]
as a wrapper, which will fetch the container token automatically, but we may
want to cache the tokens to avoid calling SCM every time, so the lower level
version in {{ContainerProtocolCalls}} might be preferred.
> Add verifier for container states
> ---------------------------------
>
> Key: HDDS-12595
> URL: https://issues.apache.org/jira/browse/HDDS-12595
> Project: Apache Ozone
> Issue Type: Sub-task
> Reporter: Ethan Rose
> Assignee: Tejaskriya
> Priority: Major
>
> Add a verifier that checks the container state in SCM for all containers in a
> replica. This can detect keys which are mapped to a container in a few bad
> states:
> * DELETING
> * DELETED
> * All replicas UNHEALTHY
> * All replicas missing
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
