[ https://issues.apache.org/jira/browse/IGNITE-3336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15347828#comment-15347828 ]
Semen Boikov commented on IGNITE-3336: -------------------------------------- So far added two system properties: IGNITE_THREAD_DUMP_ON_EXCHANGE_TIMEOUT - flag to additionaly log thread dump on hang, IGNITE_LONG_OPERATIONS_DUMP_TIMEOUT - timeout to periodically check that there are no hanging operations and log them. Changed debug logging for transaction messages to use special category (org.apache.ignite.cache.tx.msg) with several sub-categories (prepare, finish, lock, recovery) , need do the same for atomic cache, partitions exchange. > Improve Ignite troubleshooting logging > -------------------------------------- > > Key: IGNITE-3336 > URL: https://issues.apache.org/jira/browse/IGNITE-3336 > Project: Ignite > Issue Type: Task > Components: general > Reporter: Semen Boikov > Assignee: Semen Boikov > Fix For: 1.7 > > > Currently there is method IgniteKernal.dumpDebugInfo which dumps important > troubleshooting information. Currently Ignite dumps this information when > partition exchange process hangs. In addition need add some background logic > which will check that there are no hanging operations (tx, atomic updates) > and dump debug info if there are any. > In addition need improve debug logging for cache operations: e.g. if I know > id of hanging tx it should be possible to grep logs by this id and find out > all steps of tx execution, maybe need add several levels of details for tx > execution to avoid too much logging. -- This message was sent by Atlassian JIRA (v6.3.4#6332)