MySQL Bugs: #6827: server crashes if innodb_buffer_pool

Bug #6827	server crashes if innodb_buffer_pool_size set too big
Submitted:	25 Nov 2004 15:32	Modified:	25 Nov 2004 22:47
Reporter:	Aleksey Kishkin	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Server: InnoDB storage engine	Severity:	S4 (Feature request)
Version:	4.1.7	OS:	Linux (Linux)
Assigned to:	Heikki Tuuri	CPU Architecture:	Any

Description:
When I incidentely set innodb_buffer_pool_size to 2G on computer that has 512M RAM and 512M swap, mysql crashed with signal 11 (SIGSEGV).

Computer obviously cannot provide such amount of memory that mysqld demanded, but I think server must handle somehow this situation. Probably it must warn and use default low value of innodb_buffer_pool_size. 

How to repeat:
run mysql server on low memory computer with my.cnf  that contains

[mysqld]
innodb_buffer_pool_size = 2G

Aleksey,

please show the FULL output of mysqld. Do not cut anything off.

Regards,

Heikki

error log

Attachment: walrus.err (application/octet-stream, text), 2.17 KiB.

sure. I attached .err file and my.cnf. Server that I used is our standard mysql, downloaded from dev.mysql.com: 
mysql-standard-4.1.7-pc-linux-i686.tar.gz

resolve_stack_dump shows following: 

0x808af93 handle_segfault + 423
0x82d6de8 pthread_sighandler + 184
0x826cb12 ut_malloc_low + 146
0x826cb98 ut_malloc + 20
0x81e5f04 buf_pool_init + 1388
0x813b027 innobase_start_or_create_for_mysql + 1375
0x80fbb74 innobase_init__Fv + 1104
0x80f4a7f ha_init__Fv + 23
0x808c1a8 init_server_components__Fv + 808
0x808c51a main + 494
0x82deed4 __libc_start_main + 212
0x8048101 _start + 33

config file

Attachment: my.cnf (application/octet-stream, text), 38 bytes.

Walrus,

not able to allocate enough memory is a fatal error. Like it says in the .err log, we crash the server and print the stack trace to get debug information.

Regards,

Heikki

Heikki, not able to allocate enough memory is a fatal error.

Nevertheless it should not crash but return an error. MySQL then will exit gracefully.

Hi!

I am changing this to a feature request.

If we know that out-of-memory is an expected condition, and NOT a result of some unknown bug, then mysqld should exit gracefully, and not print a stack trace.

Regards,

Heikki

-------------------------------------------------
Many files:
  Fix bug #6817 and bug #6827 : do not crash the server if the buffer pool becom
es filled by the locks of ONE huge transaction, return error 1206 instead; do no
t crash the server, but return error if we cannot allocate memory for the buffer
 pool at a mysqld startup
-------------------------------------------------

If memory runs out at some other place, then we still crash mysqld to get the stack trace.

Thank you,

Heikki

Forgot to mention that the fix is in 4.1.8.

--Heikki