| Bug #52226 | Relay logs rotate inappropriately | ||
|---|---|---|---|
| Submitted: | 19 Mar 2010 16:27 | ||
| Reporter: | Shawn Green | Email Updates: | |
| Status: | Verified | Impact on me: | |
| Category: | MySQL Server: Replication | Severity: | S3 (Non-critical) |
| Version: | 5.1.41 | OS: | Any |
| Assigned to: | Assigned Account | CPU Architecture: | Any |
[19 Mar 2010 16:28]
MySQL Verification Team
The opening segments of the mysqldump readthrough of the first binary log
Attachment: bin1.txt (text/plain), 3.16 KiB.
[19 Mar 2010 16:28]
MySQL Verification Team
The entire mysqlbinlog ouput for the first relay log
Attachment: relay1.txt (text/plain), 1.58 KiB.
[19 Mar 2010 16:29]
MySQL Verification Team
The opening sections of the mysqlbinlog output of the second relay log
Attachment: relay2.txt (text/plain), 4.39 KiB.
Description: The relay logs are rotating inappropriately. Here is a list of binary logs and their sizes: 52,803,053 CCCCCC-bin.000001 56,102,876 CCCCCC-bin.000002 56,102,940 CCCCCC-bin.000003 56,102,940 CCCCCC-bin.000004 56,102,940 CCCCCC-bin.000005 And the relay logs generated by the slave 156 CCCCC-relay-bin.000001 52,429,178 CCCCC-relay-bin.000002 374,285 CCCCC-relay-bin.000003 200 CCCCC-relay-bin.000004 52,429,573 CCCCC-relay-bin.000005 3,673,713 CCCCC-relay-bin.000006 200 CCCCC-relay-bin.000007 52,428,889 CCCCC-relay-bin.000008 3,674,461 CCCCC-relay-bin.000009 200 CCCCC-relay-bin.000010 52,428,889 CCCCC-relay-bin.000011 3,674,461 CCCCC-relay-bin.000012 200 CCCCC-relay-bin.000013 52,428,889 CCCCC-relay-bin.000014 3,674,461 CCCCC-relay-bin.000015 200 CCCCC-relay-bin.000016 How to repeat: Replicate between two 5.1.41 instances. You don't need to actually process any of the relay logs, all you need to activate is the IO_THREAD and stream the binary logs to the slave. I will attach some overlapping parts of the mysqlbinlog outputs from the first binary log and the first two relay logs. Please note that relay-bin.000002 rotates backwards to relay-bin.000001 before continuing with the next statement from the actual binary log. The --max-relay-log-size setting is about double the value for --max-binlog-size See the attached files: bin1.txt - an extract from the head of the first binary log relay1.txt - the entire first relay log relay2.txt - an extract from the head of the second relay log Suggested fix: Fix the "rotate relay log" logic to behave more predictably.