On 2013.11.11 15:35, tclug at freakzilla.com wrote: > Now I know that when you just build a RAID array with mdadm it lets you > use it right away even though it's actually still building the array, and > I thought maybe ZFS does that, too, which is degrading the performance. > But zpool status doesn't show any info about that, so I'm wondering if > this kind of performance is temporary or not, and how to determine that... zpool status should tell you if it's doing something like scrubbing or resilvering. However, there isn't a whole lot to be done until you start writing data. > Second, I've had three "One or more devices has experienced an > unrecoverable error." Each has been on a different drive. I don't know how > to check /what/ the error was, though. Now, each and every one of these > drives has gone through a 3+ day badblocks and come out with > zero/zero/zero errors. Two of those errors have been on drives I've been > using in my previous mdadm array with no errors whatsoever for 6 months. > Is there any way to get a more... informative idea of what exactly error > is being encountered. zpool status should tell you what kind of error (read/write/checksum), and which files, if any, were affected. I am actually still a bit new to ZFS and have not yet had a disk fail (I've had read and checksum errors on single-disk pools that resulted in the loss of a file or two, but not had an entire disk show as faulted). It could be that ZFS is finding problems that other file systems didn't, or it could be bugs in the ZFS implementation causing data corruption or other problems (I have no idea how old the implementation you're using is or even what OS since you didn't specify). I'm not qualified to say how stable ZFS on Linux is, so I want to make it clear that I am not implying that bugs are the likely culprit, but rather a possibility. I'm currently using a single-disk ZFS-on-root pool, another single-disk pool, and a degraded 5-disk RAIDZ (NewEgg is so slow when is comes to shipping replacements!), and have not had issues, other than some performance issues which were fixed in FreeBSD 10-BETA1 (running -BETA3 right now), and some data loss on a couple of my older, less than reliable disks.