[ 
https://issues.apache.org/jira/browse/IGNITE-24804?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Roman Puchkovskiy updated IGNITE-24804:
---------------------------------------
    Description: 
ChangeIndexStatusTask waits for some condition to be satisfied (this is usually 
an await for all RW transactions in the cluster that were started on catalog 
versions earlier than some specific version to finish) and then issues a switch 
index state command to the catalog. But the table might have already been 
dropped to the moment when the index state switch is attempted, so we get an 
IndexNotFoundValidationException, which is irrelevant as the index is not there 
anymore, so we don't need to switch its state.

We should just ignore such exceptions.

Also, if we get an exception that we should not ignore, we must trigger the 
failure manager in addition to logging the exception.

> Ignore 'index not found' exception when switching index state
> -------------------------------------------------------------
>
>                 Key: IGNITE-24804
>                 URL: https://issues.apache.org/jira/browse/IGNITE-24804
>             Project: Ignite
>          Issue Type: Improvement
>            Reporter: Roman Puchkovskiy
>            Assignee: Roman Puchkovskiy
>            Priority: Major
>              Labels: ignite-3
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> ChangeIndexStatusTask waits for some condition to be satisfied (this is 
> usually an await for all RW transactions in the cluster that were started on 
> catalog versions earlier than some specific version to finish) and then 
> issues a switch index state command to the catalog. But the table might have 
> already been dropped to the moment when the index state switch is attempted, 
> so we get an IndexNotFoundValidationException, which is irrelevant as the 
> index is not there anymore, so we don't need to switch its state.
> We should just ignore such exceptions.
> Also, if we get an exception that we should not ignore, we must trigger the 
> failure manager in addition to logging the exception.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to