Response Delays

Gary Horton gary at orion.atd.ucar.edu
Wed Jun 5 04:40:00 AEST 1991


I am running on an SS1+, SunOS 4.1, Open Windows.  My SCSI bus serves two
internal 104M Quantum ProDrive disks and one external 669M SUN0669 drive,
as well as a 150M tape drive.  This workstation is an NFS server for as many
as 5 other hosts in our local network.  Pretty ordinary stuff.

My problem is this: there are frequent delays of as much as 10-15 seconds
between, e.g. asking vi to do a page down / asking unix for an ls  /
asking my application to respond, and the actual response.  In each case,
the feedback is there from, e.g. a command tool window echoing my keyboard
input, or an OL button inverting; but the action is not executed for a VERY
long time, and (not surprisingly) always AFTER hearing the gurgling noise from 
my disk indicating it has taken some needed action.  The 10-15 second delay
is much longer than it used to be or than it ought to be.  Don't you think
so? Other SS1+'s do NOT behave this way that I have seen.  

I have run

% vmstat 3

to monitor things, and the only notable events that *seem* to occur during
these delays is the number of "blocked" processes increases to some non-zero
number; this number goes back to zero immediately after the gurgling and
subsequent response-action.  I have ruled out process swapping or excessive
paging; that would be expected behavior when getting the attention of a
process that has not been bothered for a "while" ... but these delays happen
within the *same* process over very short intervals of time.  There is no
huge network traffic I can detect, at least not with etherfind.  Just the
expected login packets, NFS serving, rsh request, etc.  One of my two 
internal drives is down for now, but this is NOT the cause since the delays
had been occurring well before it bit the dust.  I am not running any
CPU hogs or inordinate number of processes, and this would show up anyway
in my vmstat CPU loads, which look pretty normal.

My question is, I hope by now, obvious: What the hell is causing this?  Has
anyone seen this before?  Was there something I heard before about the
SunOS 4.1 swapping the mouse code out?  Any clues where to look or other
net newsgroups to ask?

Please send your responses to my return address as I seldom have the time to
read the netnews.  Thanks mucho in advance...





More information about the Comp.sys.sun mailing list