You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Anton Vinogradov (Jira)" <ji...@apache.org> on 2020/04/27 16:25:00 UTC
[jira] [Created] (IGNITE-12956) Fully prepared TX rolled back on
recovery if TX coordinator failed and some primary has only reads
Anton Vinogradov created IGNITE-12956:
-----------------------------------------
Summary: Fully prepared TX rolled back on recovery if TX coordinator failed and some primary has only reads
Key: IGNITE-12956
URL: https://issues.apache.org/jira/browse/IGNITE-12956
Project: Ignite
Issue Type: Task
Reporter: Anton Vinogradov
We have 3 nodes with partitioned cache with 2 backups.
A - tx coordinator.
B and C - other nodes.
Let's start tx from A and perform
{noformat}
cache.put(primaryKeyA, someVal);
cache.get(primaryKeyB;
tx.prepare();
{noformat}
then kill A.
Expected: tx recovered and
{noformat}
cache.get(primaryKeyA)!=null
{noformat}
Actual: tx rolled back and
and
{noformat}
cache.get(primaryKeyA)==null
{noformat}
Reason: Node C has only 1 active transaction (because reads not propagated to backup), but expect to have 2.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)