[
https://issues.apache.org/jira/browse/IGNITE-23125?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Denis Chudov updated IGNITE-23125:
----------------------------------
Description:
This assertion happens when ReplicaStateManager gets an event notification
about primary replica elected, and replica is already stopped. Usually, replica
is reserved as primary at the moment when this notification is gotten (and
there is clear HB), and cannot be stopped because of assignments change if it
is reserved as primary. But it can be stopped because of replica restart (which
happens in this test) and this assertion doesn't take into account this reason.
It seems that this assertion is invalid and can be removed.
Other solution may be adding special "restarting" flag to know that the primary
is being restarted while asserting.
See ReplicaManager.ReplicaStateManager#onPrimaryElected.
was:
This assertion happens when ReplicaStateManager gets an event notification
about primary replica elected, and replica is already stopped. Usually, replica
is reserved as primary at the moment when this notification is gotten (and
there is clear HB), and cannot be stopped because of assignments change if it
is reserved as primary. But it can be stopped because of replica restart (which
happens in this test) and this assertion doesn't take into account this reason.
It seems that this assertion is invalid and can be removed.
See ReplicaManager.ReplicaStateManager#onPrimaryElected.
> AssertionError: Unexpected primary replica state STOPPED in
> ItRestartPartitionsCommandTest.testRestartAllPartitions
> -------------------------------------------------------------------------------------------------------------------
>
> Key: IGNITE-23125
> URL: https://issues.apache.org/jira/browse/IGNITE-23125
> Project: Ignite
> Issue Type: Bug
> Reporter: Denis Chudov
> Priority: Major
> Labels: ignite-3
>
> This assertion happens when ReplicaStateManager gets an event notification
> about primary replica elected, and replica is already stopped. Usually,
> replica is reserved as primary at the moment when this notification is gotten
> (and there is clear HB), and cannot be stopped because of assignments change
> if it is reserved as primary. But it can be stopped because of replica
> restart (which happens in this test) and this assertion doesn't take into
> account this reason.
> It seems that this assertion is invalid and can be removed.
> Other solution may be adding special "restarting" flag to know that the
> primary is being restarted while asserting.
> See ReplicaManager.ReplicaStateManager#onPrimaryElected.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
