Subject: [netatalk-admins] Strange client 'freezes'
From: Kadinger Andras (bandit@freeside.elte.hu)
Date: Wed Aug 26 1998 - 18:57:08 EDT
I run netatalk-1.4b2+asun2.0a18 under Linux 2.0.34.
Sometimes (and this gets more and more often recently), Macs (OS 8.1)
logged into the server tend to display the network activity arrows and
'freeze' for a couple of minutes (even the menubar clocks doesn't get
updated); at one of these occasions I happened to sit at the server and
looking at top, it displayed about 10-30% CPU usage for afpd (this is a
P-133), while I heard seeking from the HDD, much like as if I had a 'find
/' running (it's not swapping; there is 48MB RAM in the machine, and
basically afpd is the only 'real' process running). The volume in question
has about 35960 used inodes (a lot of font files).
Apparently, our other server (an ANS 700/150) gets scanned in the same
manner, but despite the similar inode usage (32xxx), IPT's optimized
desktop database or something else apparently makes the process much
faster.
Any insights intowhy OS 8.1 has the habit of scanning mounted volumes? Bad
permissions maybe make it impossible for OS 8.1 to update something
forcing to rescan the volume again and again (much like the dancing icons
problem)?
Sincerely,
Andras Kadinger
bandit@freeside.elte.hu
This archive was generated by hypermail 2b28 : Sat Dec 18 1999 - 16:33:10 EST