As a happy camper I started the scrub on my other pool, a mirrored 2*4TB.
BUT, it was not only very slow it was almost hibernating
at 600-700kB/s (kiloByte!). My old trick of export/re-import did not make any difference. OK, maybe som buffers or whatever so I rebooted the box. Still slooooow. And it skipped scrubbing and started resilvering:
- Code: Select all
pool: hera
state: ONLINE
status: One or more devices is currently being resilvered. The pool will
continue to function, possibly in a degraded state.
action: Wait for the resilver to complete.
scan: resilver in progress since Wed May 21 14:28:14 2014
1024M scanned out of 3,37T at 803K/s, (scan is slow, no estimated time)
940M resilvered, 0,03% done
config:
NAME STATE READ WRITE CKSUM
hera ONLINE 0 0 0
mirror-0 ONLINE 0 0 0
disk1s2 ONLINE 0 0 0
disk4s2 ONLINE 0 0 1 (resilvering)
logs
disk3s2 ONLINE 0 0 0
cache
disk3s3 ONLINE 0 0 0
errors: No known data errors
1. Why so slow? The number of IOPS is around 70. So even if it was all 1 byte files, the IOPS should be way higher, yes? And even with the old version, the speed was never under 10MB/s.
2. If it found a checksum error, why resilver the whole disk?
Nothing suspicious in
tail -10000f /var/system/log | grep -E '(zfs)|(disk)' what I can see.