| To: | Christian Rice <xian@xxxxxxxxxxx> |
|---|---|
| Subject: | Re: file corruption |
| From: | Chris Wedgwood <cw@xxxxxxxx> |
| Date: | Thu, 1 Apr 2004 20:44:56 -0800 |
| Cc: | Dmitry Nikiforov <dniq_kraft@xxxxxxxxxxxxxxx>, linux-xfs@xxxxxxxxxxx |
| In-reply-to: | <406CCF08.9020309@tippett.com> |
| References: | <406AF7B6.6030405@dniq-online.com> <20040402001801.GA24900@dingdong.cryptoapps.com> <406CB95B.4040500@dniq-online.com> <20040402011618.GA25511@dingdong.cryptoapps.com> <406CC518.1090204@dniq-online.com> <20040402015022.GA25936@dingdong.cryptoapps.com> <406CCF08.9020309@tippett.com> |
| Sender: | linux-xfs-bounce@xxxxxxxxxxx |
On Thu, Apr 01, 2004 at 06:25:12PM -0800, Christian Rice wrote: > Has anyone mentioned turning off write caching? > > hdparm -W 0 /dev/hda Actually, this is a good point, until XFS gets some write-barrier support this is probably a good idea. > This for me made the difference between rebuilding up to five system > disks per day (out of one hundred desktop workstations, and that was > a couple years ago), and rebuilding...none per day. Newer drives with 8MB caches are worse... you can get quite a bit of data reordered and/or lost. I'm not sure if TCQ makes this worse or not, I suspect it probably does. Turning off write-caching hurts a little in performance with once there is write-barrier support write-caching can be reenabled. --cw |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: file corruption, Ethan Benson |
|---|---|
| Next by Date: | Re: file corruption, Chris Wedgwood |
| Previous by Thread: | Re: file corruption, Stephen C. Tweedie |
| Next by Thread: | Re: file corruption, Steve Wray |
| Indexes: | [Date] [Thread] [Top] [All Lists] |