2

Can anyone advice what is going on with my setup. I have mysql server on ubuntu which periodically produce very high iowait. for long time up to 10-20 minutes. During this time system and database almost not available. Sit that using this database just hangs. I checked vmstat during this period and it shows numbers like below

 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa
 1 22      0  34712   8260 583416    0    0   660   935   76   99  6  2 84  6
 0 25      0  34560   8280 582932    0    0 42360 27008 2304 1804  9  3  0 84
 0 29      0  34560   8320 583676    0    0 41160 21524 2360 1763  4  4  0 92
 3 20      0  35912   8328 581532    0    0 12940  6856  766  764  1  0  0 99
 1 30      0  34512   8348 581804    0    0  4532  3748  925 1373  4  4  0 92

so iowait is large. I am guessing MYSQL which has 4 gb configured for innodb pool and has database of size around 6-8 gb is swapping. During this time when I used df I saw that root drive almost full it was shown 95%. When I restarted mysql in couple minutes it restarted and all came back to norm. And space on root drive (10gb) came back to 25%. I am running mysql on ebs device on amazon ec2.

What are my options? Box is 8gb large instance of ubuntu 10.4.

I will appreciate any help as I was googling and trying to solve it for couple weeks already. Thanks

user330026
  • 21
  • 2

3 Answers3

2

That's a lot of disk I/O.

You might want to check using iotop what is generating the I/O (backups?). The mysql processlist may give you (us?) further clues. Also, try running mysqltuner.pl against the dbms.

symcbean
  • 19,931
  • 1
  • 29
  • 49
0

I had a similar issue on my Cloud provider (not amazon)... I made some benchmark to check the disk performance with sysbench

The Gold Cloud was using a SAN Storage...

SRV1 (gold)
Read 0b  Written 10Gb  Total transferred 10Gb  (53.76Mb/sec)
Read 10Gb  Written 0b  Total transferred 10Gb  (22.627Mb/sec)

SRV2 (gold)
Read 0b  Written 10Gb  Total transferred 10Gb  (58.654Mb/sec)
Read 10Gb  Written 0b  Total transferred 10Gb  (21.617Mb/sec)

After that, I migrated the servers to what they call silver cloud, that uses shared SAS disks.

SRV1 (silver)
Read 0b  Written 10Gb  Total transferred 10Gb  (106.12Mb/sec)
Read 10Gb  Written 0b  Total transferred 10Gb  (78.24Mb/sec)

SRV2 (silver)
Read 0b  Written 10Gb  Total transferred 10Gb  (103.67Mb/sec)
Read 10Gb  Written 0b  Total transferred 10Gb  (98.747Mb/sec)

My mysql backup before and after:

Backup Start Time Fri Jun 24 02:20:02 BRT 2011

Backup End Fri Jun 24 03:07:01 BRT 2011

Backup Start Time Sat Jun 25 02:20:02 BRT 2011

Backup End Sat Jun 25 02:36:44 BRT 2011

Community
  • 1
Marcelo Bittencourt
  • 610
  • 4
  • 11
0

Thanks, guys, looks like with more memory problem disappeared almost, only once a day short IOWAIT. I guess answer is more memory. – user330026 just now edit

user330026
  • 21
  • 2