[
https://issues.apache.org/jira/browse/IGNITE-6433?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Pavel Kovalenko updated IGNITE-6433:
------------------------------------
Description:
If PartitionLossPolicy.IGNORE is used and we have lost some partition which
would belong to us because of affinity assignment and its state was RENTING
then we just ignore such partition and don't move to LOST state.
We should either wait for eviction asynchronously or cancel eviction for such
partition and move it's state to LOST
was:
If PartitionLossPolicy.IGNORE is used and we have lost some partition which
would belong to us because of affinity assignment and its state was RENTING
then we would wait for its eviction completing what would hang cluster (the
time of exchange would significantly increase).
Instead of waiting we should cancel eviction and it's all.
> We need to handle possible eviction when we should own a partition because we
> had lost it
> -----------------------------------------------------------------------------------------
>
> Key: IGNITE-6433
> URL: https://issues.apache.org/jira/browse/IGNITE-6433
> Project: Ignite
> Issue Type: Bug
> Components: cache
> Affects Versions: 2.1
> Reporter: Eduard Shangareev
> Priority: Critical
> Attachments: 6433-thread-dump.txt
>
>
> If PartitionLossPolicy.IGNORE is used and we have lost some partition which
> would belong to us because of affinity assignment and its state was RENTING
> then we just ignore such partition and don't move to LOST state.
> We should either wait for eviction asynchronously or cancel eviction for such
> partition and move it's state to LOST
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
