Bug #69213 A killed connection can cause failover
Submitted: 13 May 2013 9:48 Modified: 17 Dec 2013 17:44
Reporter: Daniël van Eeden (OCA) Email Updates:
Status: Closed Impact on me:
None 
Category:MySQL Utilities Severity:S3 (Non-critical)
Version:1.3.1 OS:Any
Assigned to: CPU Architecture:Any

[13 May 2013 9:48] Daniël van Eeden
Description:
When testing mysqlfailover:

I manually killed the connection from mysqlfailover to test the reconnection.

This mostly worked fine but sometimes it caused a failover to start. There was not anything obvious in the mysqlfailover log.

The log:
2013-05-13 11:12:55 AM INFO Master status: binlog: db10-bin.000004, position:5328
2013-05-13 11:13:13 AM CRITICAL Master is confirmed to be down or unreachable.
2013-05-13 11:13:13 AM INFO Failover starting in 'auto' mode...

How to repeat:
Setup replication with 5.6/GTID.

Then kill the connection from mysqlfailover a number of times.

Suggested fix:
1. Send more information to the logfile about why a failover happened:
For example: "Master didn't respond to MySQL Ping for 3 times" (Clearly indicate if it's MySQL Ping or OS Ping)

2. Make sure the defaults won't cause a failover when this isn't needed.
[8 Jul 2013 12:46] MySQL Verification Team
Hello Daniel,

Thank you for the bug report. 
Verified as described.

Thanks,
Umesh
[16 Dec 2013 19:18] Chuck Bell
Fixed in release-1.3.6
[17 Dec 2013 17:44] Philip Olson
Fixed as of the upcoming MySQL Utilities 1.3.6 / 1.4.1 releases, and here's the changelog entry:

The reconnect would sometimes fail after manually killing a connection
from the "mysqlfailover" utility.

Thank you for the bug report.