On 12/20/10 11:39 AM, Tim Evans wrote: > On 12/20/10 10:38, James Pearson wrote: >> Tim Evans wrote: >>> Seeing very slow throughput for RHEL 5.x NFS client NFS mounting a >>> Thecus N7700 NAS with "default" NFS options. >>> >>> These NAS have embedded Linux OS of some sort. >>> http://www.thecus.com/product.php?PROD_ID=14 >>> >>> dump is trickling along at about 530 KB/s; at this rate, a full dump >>> of a 150GB filesystem will take 3+ days: >>> >>> DUMP: 3.17% done at 531 kB/s, finished in 81:30 >>> >>> I've had good luck with 'proto=udp,rsize=8192,wsize=8192' on Solaris >>> systems mounting other embedded-Linux NAS (i.e., NetGear ReadyNAS), >>> but this doesn't seem to make any difference with the Thecus. >>> >>> Anyone have one of these and can suggest options to use? Thanks. >> >> What mount options are you using (i.e. from /proc/mounts) ? > > As I said (see above) pure-D defaults for the NFS mount. Changing to the > second set (above) didn't make any difference, although it did help > majorly with a Fedora 13 client and the same NAS.
Tim, We'd like to see the output of /proc/mounts to help you with the problem. "The defaults" is not good enough because they are negotiated. When you don't specify mount options, they're not necessarily going to be the most optimal. It depends on the kernel versions on both client and server. Most newer clients (depending on the server) will default to TCP, 64K rsize and wsize. /Brian/ _______________________________________________ rhelv5-list mailing list [email protected] https://www.redhat.com/mailman/listinfo/rhelv5-list
