fingerhu@ircam.fr (Michel Fingerhut) (04/22/91)
In spite of upgrading to 4.1 and of including patches as supplied by the local
dec support (incl. kern_synch), users often experience a long delay (typically
30 seconds) in the execution of most common commands on a system with a very
light load, which is totally unacceptable.
What happens is that the csh launching the commands forks a csh which then gets
stuck in a disk wait, for 35 secs. or so, as ps shows:
F UID PID PPID CP PRI NI ADDR SZ RSS WCHAN STAT TT TIME COMMAND
11008001 0 290 215 33 1 0 187f 60 48 6142c S p0 0:03 rlogind
11808001 70 295 290110 3 0 19cd 208 176 fb550 S p0 0:09 -csh (csh)
1008001 70 488 295 1 -5 0 1f4e 208 176 fbc24 D p0 0:00 -csh (csh)
The line of the "waiting csh" always appears oddly skewed.
Can ANYONE help us? We have been experiencing this since version 4.0 of
Ultrix (which was the 1st to allow more than one CPU to run), always reported
the problem, and told it was not reproducible elsewhere. Here it is.