[MDEV-4328] maria-galera: Error in Log_event::read_log_event() Found invalid event in binary log Created: 2013-03-26 Updated: 2015-03-28 Resolved: 2013-03-26 |
|
| Status: | Closed |
| Project: | MariaDB Server |
| Component/s: | None |
| Affects Version/s: | 5.5.29-galera |
| Fix Version/s: | None |
| Type: | Bug | Priority: | Major |
| Reporter: | Henrik Ingo | Assignee: | Seppo Jaakola |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | galera | ||
| Environment: |
|
||
| Attachments: |
|
| Description |
|
The following bug surfaced by running sysbench 0.5 via the attached script (test_sysbench). Note that the script will do a few other things too, such as DROP and CREATE DATABASE. This may not be significant. Sysbench is run against node 1, with a passive slave node 2. During the sysbench prepare run, one of the CREATE TABLE statements will almost always fail with the following in error log on node 2:
This is the binlog from node 1 after restart:
Note that sbtest1 is the second CREATE TABLE statement that was committed. All other tables from sbtest2 to sbtest3 have been successfully created on the slave. sbtest2 even received one insert before node shuts down due to insert to sbtest1 failing. Note that setting wsrep_debug=1 makes this issue go away. (Ie it is a usable workaround!) |
| Comments |
| Comment by Henrik Ingo [ 2013-03-26 ] |
|
The GRA files. The first one is the error, the two others are only a consequence of the table missing. |
| Comment by Henrik Ingo [ 2013-03-26 ] |
|
Elena: Just FYI, this surfaced in test run by Ivan Zoratti. I then helped him to verify it. |
| Comment by Elena Stepanova [ 2013-03-26 ] |
|
Hi Henrik, Do you happen to have the cnf file or full show variables output (not just wsrep)? |
| Comment by Henrik Ingo [ 2013-03-26 ] |
|
Attaching my.cnf and show variables. Galera version is 23.2.4 and this is an Amazon Linux AMI (see Environment). |
| Comment by Seppo Jaakola [ 2013-03-26 ] |
|
Confirmed, reproduced with MariaDB Galera Cluster 5.5.29. But, cannot reproduce with upstream Galera Cluster revisions. Here is a simple command for reproducing: sysbench --test=tests/db/parallel_prepare.lua --oltp_tables_count=100 --oltp-table-size=10 --num-threads=10 --mysql-host=127.0.0.1 --mysql-port=3306 --mysql-user=root run |
| Comment by Seppo Jaakola [ 2013-03-26 ] |
|
This bug is regression from: http://bazaar.launchpad.net/~maria-captains/maria/maria-5.5-galera/revision/3389 |
| Comment by Seppo Jaakola [ 2013-03-26 ] |
|
Fix has been pushed in revision: http://bazaar.launchpad.net/~maria-captains/maria/maria-5.5-galera/revision/3391 |
| Comment by Seppo Jaakola [ 2013-03-26 ] |
|
The bug was due to a race condition happening when several concurrent DDL statements were processed, sysbench's parallel_prepare causes such work load |