This need to be correlated to the over-the-wire traffic and the format of the backend filesystem to determine if these I/Os are spurious. I doubt that they are given that the NFS server doesn't attempt any reads before writes unless otherwise directed by the client's requests or as a side effect of the underlying filesystem.
Spencer On Jun 4, 2008, at 10:54 AM, Minos wrote: > Looking at the output from iosnoop utility you can see that nfsd is > doing some reads when it shouldn't. This log was captured while > doing a 32k write. There wasn't any other activity on the system > and nfsd was just started and freshly mounted. > > UID PID D BLOCK SIZE COMM PATHNAME > 1 1170 W 72271424 8704 nfsd /test/test1 > 1 1170 R 72271458 512 nfsd <none> > 1 1170 R 72271458 512 nfsd <none> > 1 1170 W 72271476 512 nfsd <none> > 1 1170 W 163840 512 nfsd /test/.inodes > 1 1170 W 72271458 512 nfsd <none> > 1 1170 W 72271477 5632 nfsd /test/test1 > 1 1170 R 72271441 512 nfsd <none> > 1 1170 W 72271441 512 nfsd <none> > 1 1170 R 72271441 512 nfsd <none> > 1 1170 W 72271441 512 nfsd <none> > 1 1170 W 72271442 8192 nfsd /test/test1 > 1 1170 W 72271458 512 nfsd <none> > 1 1170 W 72271459 8704 nfsd /test/test1 > 1 1170 R 72271476 512 nfsd <none> > 1 1170 W 72271476 512 nfsd <none> > 1 1170 R 72271476 512 nfsd <none> > > > This message posted from opensolaris.org > _______________________________________________ > nfs-discuss mailing list > nfs-discuss at opensolaris.org
