[
https://issues.apache.org/jira/browse/IGNITE-21500?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
 Kirill Sizov updated IGNITE-21500:
-----------------------------------
Summary: Retry implicit transactions when primary replica failed or moved
(was: Retry implicit full transactions in case of exceptions related to primary
replica failure or move)
> Retry implicit transactions when primary replica failed or moved
> ----------------------------------------------------------------
>
> Key: IGNITE-21500
> URL: https://issues.apache.org/jira/browse/IGNITE-21500
> Project: Ignite
> Issue Type: Improvement
> Reporter: Denis Chudov
> Priority: Major
> Labels: ignite-3
>
> *Motivation*
> Implicit transactions usually are "full" and include just one transactional
> request which can be safely retried in case of the primary replica related
> exceptions (PrimaryReplicaMissException, etc.). So users will never see these
> exceptions in case of primary replica failures.
> *Definition of done*
> PrimaryReplicaMissException, PrimaryReplicaAwaitException,
> TransactionExceptions with messages like "Failed to get the primary replica",
> "Failed to resolve the primary replica node" are not propagated to the users
> in cases of implicit full transactions. If it is not possible to await for
> primary replica, PrimaryReplicaAwaitException or transaction exception with
> timeout is still possible.
> *Implementation detail*
> Likely all that is required to do - is to handle the mentioned kinds of
> exceptions in {{InternalTableImpl.wrapReplicationException}} and rethrow an
> exception with the desired message.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
