1

So we have some workstations with identical hardware.

The Fedora14 box has a couple weeks uptime and still get good performance.

hdparm -tT /dev/sda
/dev/sda:
Timing cached reads:   21766 MB in  2.00 seconds = 10902.12 MB/sec
Timing buffered disk reads:  586 MB in  3.00 seconds = 195.20 MB/sec

The Cent 5.5 boxes however seem to be okay after a reboot,

/dev/sda:
 Timing cached reads:   34636 MB in  2.00 seconds = 17354.64 MB/sec
 Timing buffered disk reads:  498 MB in  3.01 seconds = 165.62 MB/sec

but some time later( unsure exactly, tested at approx 1 day uptime)

/dev/sda:
 Timing cached reads:   2132 MB in  2.00 seconds = 1064.96 MB/sec
 Timing buffered disk reads:  160 MB in  3.01 seconds =  53.16 MB/sec

drop to this. This is with very low load. I believe they all have the same bios settings. Any ideas what could cause this on Cent? Ask for more info. It might also be worth noting, that passing the --direct flag causes the slow boxes to perform similarly to the non-slow ones for buffered disk reads.

Flamewires
  • 433
  • 2
  • 10
  • Any (especially software) RAID there? Check its status, is it scrubbing or rebuilding? LVM? – Natmaka Feb 28 '11 at 17:01
  • theres a dell sas/raid controller in every box, no raid created tho. Theres also another cent box that is similar(but not identically) speced, that experiences similar slwodowns and it does have a HW raid 5, not scrubbing tho. So I doubt it is something to do with that since raid and no raid both have issues. This is happening across 2-3 cent boxes. Also no LVM on any of the boxes. – Flamewires Feb 28 '11 at 19:46
  • Check the cache policy for the RAID volumes (if read ahead is enabled, etc). – gtirloni Mar 08 '11 at 23:12

0 Answers0