Bug #70659 Make crash safe slave work with gtid + less durable settings
| Submitted: | 18 Oct 2013 18:31 | Modified: | 21 Oct 2013 13:23 |
|---|---|---|---|
| Reporter: | Yoshinori Matsunobu (OCA) | Email Updates: | |
| Status: | Verified | Impact on me: | |
| Category: | MySQL Server: Replication | Severity: | S3 (Non-critical) |
| Version: | 5.6.14 | OS: | Any |
| Assigned to: | CPU Architecture: | Any |
[18 Oct 2013 18:31]
Yoshinori Matsunobu
[21 Oct 2013 13:23]
MySQL Verification Team
Hello Yoshinori, Thank you for the bug report and test case. Verified as described. Thanks, Umesh
[17 May 2018 23:40]
Pura Vida
Regarding the suggested fix: "Truncating binary log based on the above position makes binary log and InnoDB consistent each other." It won't work well if this slave serves as a master of its own slave because the transactions in binary log may have already replicated. I would change it skip writing to the binary for transactions already recorded there.
[21 Aug 2018 17:53]
Jean-François Gagné
Related with . suggested fix for MySQL 5.7 and 8.0: Bug#92109.
[5 Dec 2020 2:05]
Jean-François Gagné
While doing tests for Bug#101874, I tested replication crash safety with GTID, sync_binlog = 0 and innodb_flush_log_at_trx_commit = 2 in a recent MySQL 8.0, and I did not get replication breakage. I suspect that replication crash safety with GTID and less durable settings was introduced with WL#9211 in MySQL 8.0.17. Could we get an official statement from Oracle on this subject ?
[7 May 2021 12:42]
Jean-François Gagné
In Bug#92109, Joe Grasse shared the following: > It looks like that [I suspect that replication crash safety with GTID and less durable settings was introduced with WL#9211 in MySQL 8.0.17.] is true, based on this section of the manual. > > If binary logging is enabled (log_bin is ON), from MySQL 8.0.17 for the InnoDB storage engine only, the server updates the mysql.gtid_executed table in the same way as when binary logging or replica update logging is disabled, storing the GTID for each transaction at transaction commit time. > > https://dev.mysql.com/doc/refman/8.0/en/replication-gtids-concepts.html So it looks like this bug is fixed in MySQL 8.0.17. In Bug#92109, I asked for this fix to be back-ported in MySQL 5.7.
[4 Aug 22:31]
Jean-François Gagné
It looks like WL#9211 [1] fixed this in MySQL 8.0.17. See the [2] blog post for detail. [1]: https://dev.mysql.com/worklog/task/?id=9211 [2]: https://jfg-mysql.blogspot.com/2026/08/mysql-8017-gtid-crash-safety-improvement.html So I think this bug can be closed, but this should be left to Oracle as I might be missing details on this subject.
Portions of this website are copyright © 2001, 2002 The PHP Group
Page generated in 0.060 sec. using MySQL 8.4.7-u6-cloud
Timestamp references displayed by the system are UTC.
Content reproduced on this site is the property of the
respective copyright holders.
It is not reviewed in advance
by Oracle and does not necessarily represent the opinion of
Oracle or any other party.