Brian,

One quick thought is that the quota initialization walks the inode file rather than walking the directory structure (because that way is usually faster AND much easier to get right). Thus the speed of quota initialization is proportional to the number of total inodes (8.7 million in your case) and not to the number of allocated inodes (1161 in your case). If your experiences on 4.2a were with filesystems with fewer inodes that would explain the difference.

Also, I'm curious, has this file system ever been populated or is it a newly created file system?

John Edwards Network Appliance

On Tue, 3 Nov 1998, Brian Tao wrote:

Is it just me, or is quota initialization on an F230 running 5.1.2

excruciatingly slow? I don't remember our other F230's being this slow at redoing tree quotas with 4.2a, especially when the filer is basically empty (only the /etc directory is populated). It's been 15 minutes since a "quota on", and the filer only reports 39% done. I'm sure it was faster in 4.2a.

# rsh toast quota vol0: quotas are initializing (39% done).

# rsh toast df Filesystem kbytes used avail capacity Mounted on /vol/vol0/ 33007568 1212560 31795008 4% /vol/vol0/ /vol/vol0/.snapshot 11002520 0 11002520 0% /vol/vol0/.snapshot

# rsh toast df -i Filesystem iused ifree %iused Mounted on /vol/vol0/ 1161 8709169 0% /vol/vol0/

# rsh toast sysstat 1 CPU NFS CIFS HTTP Net kB/s Disk kB/s Tape kB/s Cache in out read write read write age 37% 0 0 0 0 0 1211 0 0 0 6 72% 0 0 0 1 0 76 0 0 0 6 22% 0 0 0 0 0 244 0 0 0 6 30% 0 0 0 1 0 1234 0 0 0 6 79% 0 0 0 2 0 56 0 0 0 6 ^C

-- Brian Tao (BT300, taob@risc.org) "Though this be madness, yet there is method in't"