burn-in zfs pool?

[wkearney99]

What's the best way to really beat the hell out of a new ZFS pool? The idea being to really hammer on the pool as much as possible. I want to burn-in a 6 drive raidz2 pool as quickly and as mercilessly as can be done. I don't need to use the machine for anything else in the interim.

These are 2tb Hitachi drives running off an Areca 1260 in JBOD mode.

 

[drescherjm]

At work for the last 2 weeks or so I have been beating up a btrfs on top of linux software raid 6 on 13 x 8 year old 250 GB SATA 1 drives (some with 100s of reallocated sectors). My method is to copy a ~300 GB data set (size of the user public folders) into multiple subvolumes until the disk was nearly full. Then hot pull a drive letting the array automatically pull in a spare then do a scrub. Then hot add the pulled disk and repeat with a different hot pulled disk..

 

[madrebel]

run iometer against the pool or run bonnie++ in a loop.

 

[wkearney99]

Creating the iobw.tst file takes eff'in FOREVER on a 6 drive raidz2! Like 8+ hours and counting before I had to do a reboot due to other concerns.

Anyone used iometer with a local dynamo instead of remote? Any advantages of doing this or not? Realizing, of course, that it would then only give local performance measures, not those through the network. Which, when testing, is sometimes a desirable option.

I managed to get it to compile and run on OI. I had to tweak the IOCommon.h to #define NOMINMAX after IOstreams (which I'm sure is probably not the ideal way to do it, but hey, it compiled).