when mysqld dies it leaves SST scripts running
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
MySQL patches by Codership |
New
|
Undecided
|
Unassigned | ||
Percona XtraDB Cluster moved to https://jira.percona.com/projects/PXC |
New
|
Undecided
|
Unassigned |
Bug Description
I still observe cases when mysqld dies, but SST script and external utils are continuing.
i.e.
ps auxww
mysql 32618 0.0 0.0 106148 1348 ? S 15:19 0:00 /bin/bash -ue /usr/local/
mysql 32631 0.0 0.0 7608 508 ? S 15:19 0:00 nc -dl 4444
mysql 32632 0.0 0.0 113948 936 ? S 15:19 0:00 tar xfi - -C /mnt/data/mysql/
but mysqld died:
===
121108 15:20:53 [Note] WSREP: Created page /mnt/data/
121108 15:21:19 [ERROR] WSREP: We ran out of resources, seemingly because we've been unsuccessfully requesting state transfer for over 116 seconds. Please check that there is at least one fully synced member in the group. Application must be restarted.
121108 15:21:19 [ERROR] WSREP: State transfer request failed unrecoverably: 35 (Resource deadlock avoided). Most likely it is due to inability to communicate with the cluster primary component. Restart required.
121108 15:21:19 [Note] WSREP: Closing send monitor...
121108 15:21:19 [Note] WSREP: Closed send monitor.
121108 15:21:19 [Note] WSREP: gcomm: terminating thread
121108 15:21:19 [Note] WSREP: gcomm: joining thread
121108 15:21:19 [Note] WSREP: gcomm: closing backend
121108 15:21:20 [Note] WSREP: view(view_
} joined {
} left {
} partitioned {
})
121108 15:21:20 [Note] WSREP: New COMPONENT: primary = no, bootstrap = no, my_idx = 0, memb_num = 1
121108 15:21:20 [Note] WSREP: view((empty))
121108 15:21:20 [Note] WSREP: gcomm: closed
121108 15:21:20 [Note] WSREP: Flow-control interval: [16, 16]
121108 15:21:20 [Note] WSREP: Received NON-PRIMARY.
121108 15:21:20 [Note] WSREP: Shifting PRIMARY -> OPEN (TO: 17556706)
121108 15:21:20 [Note] WSREP: Received self-leave message.
121108 15:21:20 [Note] WSREP: Flow-control interval: [0, 0]
121108 15:21:20 [Note] WSREP: Received SELF-LEAVE. Closing connection.
121108 15:21:20 [Note] WSREP: Shifting OPEN -> CLOSED (TO: 17556706)
121108 15:21:20 [Note] WSREP: RECV thread exiting 0: Success
121108 15:21:20 [Note] WSREP: recv_thread() joined.
121108 15:21:20 [Note] WSREP: Closing slave action queue.
121108 15:21:20 [Note] WSREP: /usr/local/
121108 15:21:20 mysqld_safe mysqld from pid file /mnt/data/
===
It would be good to have a way to kill SST script and nc + tar utils, as they are using network IP address and port,
and it is impossible to start node again without killing them manually.
in fact, even I do clean mysqladmin shutdown,
if donor was running at that time, I still see running processes Percona- XtraDB- Cluster- 5.5.28- 23.7.366. Linux.x86_ 64//bin/ wsrep_sst_ xtrabackup --role donor --address 10.4.101. 225:4444/ xtrabackup_ sst --auth (null) --socket /tmp/mysql.sock --datadir /mnt/data/mysql/ --defaults-file /etc/my.cnf --gtid 9db8f201- 2858-11e2- 0800-0532898e05 9e:19606001 innobackupex --galera-info --tmpdir=/tmp --stream=tar --defaults- file=/etc/ my.cnf --socket= /tmp/mysql. sock /tmp file='/ etc/my. cnf' --socket= '/tmp/mysql. sock' --unbuffered -- >/tmp/jq644fvuFh 2>/tmp/FfogPFCicM file=/etc/ my.cnf --socket= /tmp/mysql. sock --unbuffered --
mysql 21920 0.0 0.0 106092 1328 pts/2 S 17:05 0:00 /bin/bash -ue /usr/local/
mysql 21932 0.1 0.0 143228 10824 pts/2 S 17:05 0:00 perl /usr/bin/
mysql 21934 0.0 0.0 106092 1160 pts/2 S 17:05 0:00 sh -c mysql --defaults-
mysql 21935 0.0 0.0 197916 1888 pts/2 S 17:05 0:00 mysql --defaults-
after mysqld was destroyed.