Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On 06/01/2012 10:37 AM, Lionel Sausin wrote: > Looks like you're using adaptative syncer rate, so verify will be > throttled when applications request I/O. Which isn't the problem here. These are two completely idle boxes set aside specifically for testing. My iostat results on the underlying device show DRBD is flooding the IO channel with as many reads as it can handle (7000-8000 per second). CPU is also very low, no more than 20% on each side. Sar says io wait is over 10%, which is actually pretty high. But like I said, in our stage and production systems which have much more capable RAID1+0's and a similarly speedy FusionIO ioDrive, we don't notice this. Partially because these kinds of devices can handle random IO much better than a 2-disk RAID-0. But my question: why is DRBD doing random IO on the verify? If it's following the map to the disk, that's a seek for each map entry. I would have thought it would take a gulp of several sequential blocks, find the map entries, calculate the MD5's, and then compare those across the network. From observation, that doesn't seem to be what's happening on either 8.3 or 8.4. -- Shaun Thomas OptionsHouse | 141 W. Jackson Blvd. | Suite 500 | Chicago IL, 60604 312-444-8534 sthomas at optionshouse.com ______________________________________________ See http://www.peak6.com/email_disclaimer/ for terms and conditions related to this email