Bug #41476 | Errors on shutdown opening ndb_binlog_index table | ||
---|---|---|---|
Submitted: | 15 Dec 2008 15:29 | Modified: | 1 Jan 2011 13:25 |
Reporter: | Geert Vanderkelen | Email Updates: | |
Status: | No Feedback | Impact on me: | |
Category: | MySQL Cluster: Replication | Severity: | S3 (Non-critical) |
Version: | mysql-5.1-telco-7.0 | OS: | Any |
Assigned to: | Assigned Account | CPU Architecture: | Any |
[15 Dec 2008 15:29]
Geert Vanderkelen
[27 Jan 2009 12:13]
Jonas Oreland
not urgent, but annoying see issue for details (or ask)
[3 Feb 2009 10:01]
Tomas Ulin
So what happens is that on server shutdown the injector thread continues to run until it has inserted all epochs into the binlog that has contents that originate from the same server. I.e. tries to make sure all "local" updates end up in the "local" binlog. However this printout shows that this does not succeed... the writing is interrupted by the server.
[14 May 2010 8:40]
Geert Vanderkelen
Always repeatable: 1) issue a LOCK mysql.ndb_binlog_index WRITE in a session, so the table is not written anymore by the injector thread 2) now, do lots of updates using a NDB API application for example and leave it running for a while. 3) shutdown the mysqld (send SIGTERM) mysqld will always report 'killed' and that it was unable to lock the ndb_binlog_index table. Tomas mentioned it earlier, this is because the rows that still need to be logged in ndb_binlog_index could not be saved in a timely manner (or at all). It's more than an annoyance because we have now more entries in the binary logs than ndb_binlog_index makes us believe.
[1 Dec 2010 13:25]
Martin Skold
Unable to reproduce, need feedback if fix for Bug#55909 ndb_binlog_index missing entry on normal mysqld shutdown removed the problem?
[2 Jan 2011 0:00]
Bugs System
No feedback was provided for this bug for over a month, so it is being suspended automatically. If you are able to provide the information that was originally requested, please do so and change the status of the bug back to "Open".