Execution of stored procedures with subqueries leaks memory
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
MySQL Server |
Unknown
|
Unknown
|
|||
Percona Server moved to https://jira.percona.com/projects/PS |
Fix Released
|
High
|
Vlad Lesin | ||
5.1 |
Invalid
|
Undecided
|
Unassigned | ||
5.5 |
Invalid
|
Undecided
|
Unassigned | ||
5.6 |
Fix Released
|
High
|
Vlad Lesin |
Bug Description
We have this problem:
with 50GB innodb buffer pool the server eats about 56GB of memory (RES) with the production workload – that’s OK. But then the memory consumption is increasing and after 3 days of runtime, it uses about 76GB of memory (RES) and the machine starts swapping out. I have to restart mysqld every 3 days.
We have the exact version of PerconaServer on the slave (without production workload, just the slave thread) and it doesn’t have this problem (54 GB mem RES after 7 days).
This problem is the same for a few recent versions of Percona server, not just the last one.
Server gets just normal OLTP traffic + mysqldump and xtrabackup nightly.
We DO use a few InnoDB tables with compression and we use some MyISAM tables with range partitioning.
TokuDB is not installed.
-- -------
mysql> select version();
+-----------------+
| version() |
+-----------------+
| 5.6.21-69.0-log |
+-----------------+
CentOS release 6.4 (Final)
Linux xxxxx 2.6.32-
my.cnf:
[mysqld]
user=mysql
datadir=
tmpdir=
innodb_
innodb_
log-error=
character-
collation_
init_connect='SET NAMES utf8'
sql-mode=
query_cache_type=0
local_infile=0
lock_wait_timeout=5
skip-name-resolve
explicit_
max_binlog_
expire_logs_days=3
binlog-format=ROW
binlog_
log-slave-updates
binlog_
binlog-
replicate-
replicate-
slave_type_
sync_binlog=1
sync_relay_log=100
relay-log-
relay_log_
max_connections
max_connect_
back_log=3000
thread_
max_allowed_
myisam-
delay-key-write=OFF
table_open_
table_definitio
loose-innodb=FORCE
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
innodb_
tmp_table_size=100M
max_heap_
event_scheduler=1
server-id=1
log-bin=
master-
relay-log=
relay-log-
Changed in percona-server: | |
status: | Incomplete → New |
tags: | added: i48579 |
tags: | added: upstream |
summary: |
- Possible memory leak + Execution of stored procedures with subqueries leaks memory |
Thank you for the problem report. Please, send the outputs of:
show global status;
show engine innodb status\G
Also I wonder if memory use changes notably after you execute FLUSH TABLES statement.