[zfs-discuss] scrub hung?
juice.qr at gmail.com
Thu Nov 29 09:44:02 EST 2012
Maybe dedup is the culprit. My dedup ratio is currently around 6.28x.
- Chris Clifton
On Thu, Nov 29, 2012 at 8:40 AM, Kate Ward <kate.ward at forestent.com> wrote:
> How many of you reporting slow stats are using deduplication? I had
> performance issues (including scrub performance) in the past when I used
> it, especially before I added an SSD (but still somewhat even after that).
> Chris at least is using deduplication without an SSD, so I'm curious what
> a 'zpool iostat -v 15' looks like during the scrub.
> - kate
> On Thu, Nov 29, 2012 at 2:27 PM, Okky <mawcikurl at gmail.com> wrote:
>> Just want to add the statistic, my data pool is a stripe of dual 2-way
>> mirror of 2 TB disks, total capacity is 3.62 TB, data are at 81%. Scrub ran
>> about 7.5 hours. Around 40% of the data are big multimedia files, 20% of it
>> are compressed VirtualBox images.
>> Sent with Sparrow <http://www.sparrowmailapp.com/?sig>
>> On Thursday, November 29, 2012 at 18:49, Aneurin Price wrote:
>> On 29 November 2012 05:56, RB <aoz.syn at gmail.com> wrote:
>> I think you both have something odd. My 1TB mirror (494GB used,
>> connected via SATA-II) takes ~2.5 hours to scrub, my 12TB RAIDZ (2TB
>> used, fast machine) takes about 25 minutes, and my 32TB dual RAIDZ-2
>> (19TB used, quad SAS-2 link to the enclosure) takes about 28 hours.
>> None of them take a long time to finish, but they do all take a little
>> bit to spin up to full speed.
>> Might as well add an extra data point:
>> My ~5TB pool (3 vdevs, each 2-way mirrored) takes around 2-3 days to
>> scrub. The beginning (maybe 5% or so) takes about a day; the next
>> 85-90% takes anywhere from 8 hours to a day, and then the remainder an
>> additional day.
>> Obviously those numbers are very, very rough.
> Kate Ward <kate.ward at forestent.com>
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the zfs-discuss