[CentOS] kswapd taking 100% cpu with no swap on system

Lamar Owen lowen at pari.edu
Mon May 9 17:51:23 UTC 2011


On Monday, May 09, 2011 01:11:17 PM Ali Ahsan wrote:
> sd 0:0:0:0: Attached scsi disk sda
>    Vendor: ATA       Model: WDC WD10EARS-003  Rev: 80.0
>    Type:   Direct-Access                      ANSI SCSI revision: 05

Are your two drives in a RAID?  cat /proc/mdstat

The WD10EARS drives are known to have some issues, having to do with both powersave and TLER (google for WDTLER), as well as 4k sector size.  I have seen these issues, but that was with Fedora, and it showed up as high system load, but not as kswapd at a high percentage.  Also google for WD10EARS linux percentage; the top result on that page is a Western Digital community forum post titled 'WD10EARS slow, slow, slow, slow - Western Digital Community'

If you install the sysstat package, you can use iostat to see if iowaits are your problem; I've used 'iostat -x 1' (and then set my konsole to wider.....) and monitor the await column for each device; pinned down a WD15EADS drive creating iowaits....

Hope that helps.



More information about the CentOS mailing list