[
https://issues.apache.org/jira/browse/IGNITE-6324?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Stanilovsky Evgeny updated IGNITE-6324:
---------------------------------------
Release Note: Before this fix it was possible to obtain persistent data in
inconsistent state. For example: 2 nodes with persistence and 0 backups,
transaction (tx) update data cross nodes (all nodes are involved into tx
update) and in the middle of commiting phase something happens with bare metal
or possibly jvm crash raises on one of existing nodes, second one will detect
it and rollback process will be triggered, some time later crashed node will be
fixed and returned into cluster after recovery from wal it would be possible to
observe data from rolled back tx. Now this situation is still possible if
there is no -DIGNITE_WAL_LOG_TX_RECORDS=true appropriate flag, i plan to
trigger one more ticket for changing this flag defaults. (was: Before this fix
it was possible to obtain persistent data in inconsistent state. For example: 2
nodes with persistence and 0 backups, transaction (tx) update data cross nodes
(all nodes are involved into tx update) and in the middle of commiting phase
something happens with bare metal or possibly jvm crash raises on one of
existing nodes, second one will detect it and rollback process will be
triggered, some time later crashed node will be fixed and returned into cluster
after recovery from wal it would be possible to observe data from rolled back
tx. )
> Transactional cache data partially available after crash and further recovery.
> ------------------------------------------------------------------------------
>
> Key: IGNITE-6324
> URL: https://issues.apache.org/jira/browse/IGNITE-6324
> Project: Ignite
> Issue Type: Bug
> Components: persistence
> Affects Versions: 2.10
> Reporter: Stanilovsky Evgeny
> Assignee: Stanilovsky Evgeny
> Priority: Critical
> Fix For: 2.11
>
> Attachments: InterruptCommitedThreadTest.java
>
> Time Spent: 50m
> Remaining Estimate: 0h
>
> If InterruptedException raise in client code during pds store operations we
> can obtain inconsistent cache after restart.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
