[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Summary of hangind HP proliant

Quick summary of the Issue I posted a few weeks go about a HP DL380G3
server running RHAS21 that was hanging under heavy disk IO. Server is
2x2.8GB hyper threading Xeons, 6GB 400Mhz RAM.

It has a built in SmartArray 5i card that was mirroring 2 disks for
root, and raid5-ing 4 disk for /apps_data.
A multi GB tar could make the box unresponsive for several seconds.
Through put to both volumes measured about the same ~20-25MB/s, using dd

We evaled a $1300 SmartArray 5312 RAID card and only got about 30MB/s
out of it, and though the hang was less obtrusive, it still was bad if
you got one dd and a tar going at the same time. We got the feeling that
the cpu section of the box was so fast as to swamp the disk IO section,
and the OS would effectively pause while till the buffers cleared to a
"reasonable" level.( cpu at this hang time would be about 90-95% idle,
and load would go from about 40-100)

We then changed the 5i config to present each disk as a raid 0 device
(no processing on the card now) and used Linux Raid to do the mirroring
and raid5-ing. We now get about 100-105MB/s thoughput to the SW raid 5,
and no more apparent hangs.

If anyone is using a SmartArray device, you may want to experiment with
SW raid instead. 

Dan Liebster

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]