Tuesday, September 14, 2010

iostat

One of the client's news portal has a surged of visitors due to a very interesting local news. The traffic doubled up overnight and the server was struggling to cope with it. The specs of the server is as below:-

Dual Xeon 5405
16GB DDR2 ECC RAM
2 x 320GB SATA 7200 (s/w RAID1)

I have tuned the webserver and mysql to allow more connections but the load was too high. The average load can go up as high as 40 during peak hours. The webserver can accept up to 352 concurrent connections now, up from 152 before the tuning. Memory usage should be still fine as only about 30MB swap being used as shown from "top" and "free".

The output from iostat is as below:-

Linux 2.6.26-2-amd64 (zz) 09/15/2010 _x86_64_

avg-cpu: %user %nice %system %iowait %steal %idle
23.95 0.00 4.44 0.80 0.00 70.82

Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
sda 28.06 1000.72 1154.44 1459026910 1683138720
sda1 28.05 992.46 1154.38 1446969730 1683053488
sda2 0.00 0.00 0.00 4 0
sda5 0.01 8.27 0.06 12056864 85232
sdb 30.07 1078.20 1154.44 1571984044 1683138720
sdb1 30.04 1069.93 1154.38 1559927008 1683053488
sdb2 0.00 0.00 0.00 4 0
sdb5 0.04 8.27 0.06 12056720 85232
md0 163.84 1221.37 1143.97 1780718146 1667876624
md1 0.01 0.01 0.05 16216 79440

Is this bad? or normal for a 3 years old server and hard disk?

Is a hardware upgrade an imminent solution for this problem? I'm thinking of an upgrade to the latest Nehalem with SAS (Hardware RAID 10), what do you think?

1 comment:

  1. its not bad at all, i mean the CPU load is 23.95 not too much, can go low :P

    however all BIG 320GB SATA most of the times
    get bad sectors so you may wanna keep an eye on them.

    ReplyDelete