You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Vladimir Ozerov (JIRA)" <ji...@apache.org> on 2019/03/11 06:33:05 UTC
[jira] [Updated] (IGNITE-10185) TX can hang forever if any runtime
exception occurs on txFinish.
[ https://issues.apache.org/jira/browse/IGNITE-10185?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Vladimir Ozerov updated IGNITE-10185:
-------------------------------------
Labels: iep-14 transactions (was: iep-14 mvcc_stabilization_stage_1 transactions)
> TX can hang forever if any runtime exception occurs on txFinish.
> ----------------------------------------------------------------
>
> Key: IGNITE-10185
> URL: https://issues.apache.org/jira/browse/IGNITE-10185
> Project: Ignite
> Issue Type: Bug
> Components: cache, mvcc
> Reporter: Andrew Mashenkov
> Priority: Major
> Labels: iep-14, transactions
> Fix For: 2.8
>
>
> The issue relates to incorrect IOOM handling that can occurs on Tx prepare\commit\rollback and can be reproduced if persistence enabled and Tx state logging into WAL enabled.
> This affects MVCC tx as it always log it's state into WAL and non-MVCC Tx with enabled WAL logging via setting IGNITE_WAL_LOG_TX_RECORDS system property.
> We have to check and fix if tx finish methods handle RuntimeExceptions in proper way.
> Good start is to force throw RuntimeException from tm().mvccPrepare() and tm().mvccFinish() methods, and check if DhtFinishFuture done correctly with exception rather then (re)throwing exception bypassing failure handler.
> The goal is to make IoomFailureHandlerTest passed after runtime failures during Tx commit\prepare\rollback.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)