shmget error

From: Gianluca Interlandi (gianluca_at_bioc.unizh.ch)
Date: Mon Jun 20 2005 - 07:46:28 CDT

Dear list,

We run NAMD on a cluster of 128 dual node AMD opterons with 1GB memory per
node. The communication is over gigabit ethernet. NAMD was compiled with
mpich-1.2.6.

NAMD produces the following error at runtime:

running /home/bioc/gianluca/NAMD/bin-gb/namd2 on 12 LINUX ch_p4 processors
Created /scratch/curcio/testgb_pid14872/PI14957
    p4_error: latest msg from perror: No space left on device
p0_15259: p4_error: OOPS: shmget failed: -1

The system administrators told us that this is a mpich bug, because mpich
doesn't check how much shared memory there is present on a node. If the
programm requires more than it is available, mpich aborts with "shmget
failed". A suggested workaround is to limit the memory requested by NAMD
untill it runs.

How can this be done?

Many thanks,

          Gianluca Interlandi

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 05:18:50 CST