> Until this thread I had always assumed it was because the read operations > were having to seek all around the original disks, not that writes were > being a pain. If you actually do end up debugging it I'd be interested > in the results.
Thanks to everyone for the input. I will hopefully find some time to try out the various suggestions next week and will report back what I find, esp if I discover how to improve these kinds of operations. The operation that inspired this post finished over night. The trace of its output is below. To briefly answer a few questions people have posed: I used "mv" between NFS mounted filesystems. The XRaid (destination) uses a fibre channel connected RAID-5 array (4 TB usable) which is managed by an XServe (hence my expectation that the destination FS was HFS). It took a little over 1.5 days wall clock to move 1.3 million files and 100 GB, and a little under 3 hours CPU time on the destination system (I started the move from the server hosting the destination file system). Ian root# tick=`date +%s`; date; time mv [23]* /Volumes/MIXRAID/ijstokes/mr/phaser/temp/; date; tock=`date +%s`; echo $(( $tock - $tick )) Wed Jul 21 16:22:39 EDT 2010 mv: 3fze/group/results/._3fze-partial-20100208.dgraph: could not copy extended attributes to /Volumes/MIXRAID/ijstokes/mr/phaser/temp/3fze/group/results/._3fze-partial-20100208.dgraph: Operation not permitted mv: 3fze/group/results/3fze-partial-20100208.dgraph/Contents/._PkgInfo: could not copy extended attributes to /Volumes/MIXRAID/ijstokes/mr/phaser/temp/3fze/group/results/3fze-partial-20100208.dgraph/Contents/._PkgInfo: Operation not permitted real 2263m18.835s user 3m37.100s sys 154m26.706s Fri Jul 23 06:05:58 EDT 2010 135799
<<attachment: ijstokes.vcf>>
_______________________________________________ bblisa mailing list [email protected] http://www.bblisa.org/mailman/listinfo/bblisa
