Monitor dies with "Too many open files" error

[Karl Nissen]

Morning,

We have recently upgraded our GRASS computer system from Solaris 2.5 to 2.5.1 o
n 
Sparc, and now seem to have a problem with the monitor processes dying in a 
random way with a "Too many open files" error. This problem was not seen with 
Solaris 2.5

This is normally a UNIX error message, indicating that there are insufficient 
file descriptors available. I have increased this number from 128 to 512 on the
 
system (see below for system descriptions), but this hasn't made any difference
.

Similarly a recompilation of the GRASS source has not corrected the problem 
either.

The only other difference that I can think of is that the GRASS system now a 
faster computer and there is some timing issue here that was not seen on the 
slower Sparc 5 that used to run this software.

The installation is somewhat non standard in that the GRASS binary is not 
installed as a standalone application. The binary is NFS mounted off a shared 
disk to the other computers that use grass for processing.

The new installation uses an Ultrasparc as the server and Sparc 4s as "clients"
, 
the old system used a Sparc 5 as the server and Sparc IPXs as clients. As 
mentioned above each user has had the file descriptor limit increased to 512.

Investigation of the problem using 'lsof' (a pubic domain file list generator) 
indicates that the XDRIVER opens up four files when started from d.mon.

At some time during the operation of the system the XDRIVER process starts 
opening further files at the rate of about one per second. These files are not 
closed, with the eventual failure of the monitor with a "too many files open" 
error.

Does anyone have any ideas on what might be causing this problem?

Karl



--------------------------------------------------------------------
Karl Nissen                            Phone:  (+61 6) 249 4613 
Department of Geography, SRMES         Fax:    (+61 6) 249 3770
The Australian National University     e-mail: karl at dart.anu.edu.au       
--------------------------------------------------------------------