pt-kill not killing connections when execute-command is set

[Tom Krouper]

While testing pt-kill I discovered that for some reason, it appears that the "execute-command" is causing the "kill" option to fail. Simple remove the execute-command option and everything works as expected. Below is an example I created to test. and show the issue.

Configuration: .pt-kill.conf
busy-time=30s
log=/root/pt_kill.log
pid=/root/pt_kill.pid
execute-command=(date >> /root/execute.log)
match-info=^SELECT
kill
print

MySQL Version:
mysqld Ver 5.6.24-72.2 for Linux on x86_64 (Percona Server (GPL), Release 72.2, Revision 8d0f85b)

Percona Toolkit Version:
pt-kill 2.2.14

Queries run and output:
[root@tslls ~]# date; mysql --comments -e "SELECT @@server_id, SLEEP(65) /* First Query */;"; date
Sat May 30 23:10:28 UTC 2015
ERROR 2013 (HY000) at line 1: Lost connection to MySQL server during query
Sat May 30 23:11:02 UTC 2015
[root@tslls ~]# date; mysql --comments -e "SELECT @@server_id, SLEEP(65) /* Second Query */;"; date
Sat May 30 23:11:43 UTC 2015
+-------------+-----------+
| @@server_id | SLEEP(65) |
+-------------+-----------+
| 0 | 0 |
+-------------+-----------+
Sat May 30 23:12:48 UTC 2015
[root@tslls ~]# touch /tmp/pt-kill-sentinel

execute-command output:
[root@tslls ~]# cat execute.log
Sat May 30 23:11:02 UTC 2015
Sat May 30 23:12:20 UTC 2015
Sat May 30 23:12:39 UTC 2015

[Gillian Gunson (ggunson)]

I was able to repeat this using Tom's test case. It looks like pt-kill will succeed on killing the first query, but all subsequent queries it will fail to kill on "MySQL server has gone away" at /usr/bin/pt-kill line 6806. And it looks like it won't reestablish the database connection until the next "SHOW FULL PROCESSLIST" fails after the two failed kill attempts.

[Nilnandan Joshi (nilnandan-joshi)]

Verified with 2.2.14.

nilnandan@desktop:~$ pt-show-grants --version
pt-show-grants 2.2.14
nilnandan@desktop:~$

nilnandan@desktop:~$ cat .pt-kill.conf
busy-time=30s
log=/home/nilnandan/pt_kill.log
pid=/home/nilnandan/pt_kill.pid
execute-command=(date >> /home/nilnandan/execute.log)
match-info=^SELECT
kill
print
nilnandan@desktop:~$

nilnandan@desktop:~$ date; mysql --comments -e "SELECT @@server_id, SLEEP(65) /* First Query */;"; date
Mon Jun 1 15:22:02 IST 2015
ERROR 2013 (HY000) at line 1: Lost connection to MySQL server during query
Mon Jun 1 15:22:33 IST 2015
nilnandan@desktop:~$

nilnandan@desktop:~$ cat .pt-kill.conf
busy-time=30s
log=/home/nilnandan/pt_kill.log
pid=/home/nilnandan/pt_kill.pid
#execute-command=(date >> /home/nilnandan/execute.log)
match-info=^SELECT
kill
print
nilnandan@desktop:~$ date; mysql --comments -e "SELECT @@server_id, SLEEP(65) /* First Query */;"; date
Mon Jun 1 15:23:45 IST 2015
+-------------+-----------+
| @@server_id | SLEEP(65) |
+-------------+-----------+
| 10000 | 0 |
+-------------+-----------+
Mon Jun 1 15:24:50 IST 2015
nilnandan@desktop:~$
nilnandan@desktop:~$ cat execute.log
Mon Jun 1 15:22:33 IST 2015
Mon Jun 1 15:24:40 IST 2015
nilnandan@desktop:~$

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PT-1285