[
https://issues.apache.org/jira/browse/YARN-5630?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15484465#comment-15484465
]
Jason Lowe commented on YARN-5630:
----------------------------------
Thanks for the +1, Arun! I'll commit this later today if there are no
objections.
Regarding the key removal, yes if we read the keys out of the store itself then
that totally makes sense. Actually I was assuming that during the rollback
when the old software saw a key that was ignored it would just remove it as it
iterated the database. I suppose we could have the entry flag whether or not
the ignored key should be preserved as long as the container is active in case
there is a subsequent roll-forward. Then there's no need for a
prepare-for-rollback once the support for the key table is in the old software
version and all we need to do is delete keys. Of course if there's more to do
than just delete keys or fail containers then something needs to be run on the
new software before downgrading to the old one.
> NM fails to start after downgrade from 2.8 to 2.7
> -------------------------------------------------
>
> Key: YARN-5630
> URL: https://issues.apache.org/jira/browse/YARN-5630
> Project: Hadoop YARN
> Issue Type: Bug
> Components: nodemanager
> Reporter: Jason Lowe
> Assignee: Jason Lowe
> Priority: Blocker
> Attachments: YARN-5630.001.patch, YARN-5630.002.patch
>
>
> A downgrade from 2.8 to 2.7 causes nodemanagers to fail to start due to an
> unrecognized "version" container key on startup. This breaks downgrades from
> 2.8 to 2.7.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
