Newsgroups: comp.parallel.pvm
From: tippetjm@esvx19.es.dupont.com
Subject: Problem with PVM_ARCH=SGIMP
Organization: DuPont (Opinions are those of the writer only)
Date: Thu, 4 Aug 1994 17:34:16 GMT
Message-ID: <1994Aug4.173416.15794@es.dupont.com>

I am trying to use PVM on a 4 processor SGI Challenge workstation running
IRIX 5.2 ( PVM_ARCH=SGIMP ), so far unsuccessfully. The following pvml.uid 
files from several attempts to run the fspmd example included with the pvm3.3 
distribution indicate the type of problem I am having. 

[t80040000] ready  3.3.0   Mon Jul 25 08:41:59 1994
[t80040000] removeshm: sem key = 40004: No such file or directory
[t80040000] removeshm: sem key = 40005: No such file or directory
[t80040000] removeshm: shm key = 4000a: No such file or directory
[t80040000] removeshm: shm key = 4000b: No such file or directory
[t80040000] removeshm: shm key = 4000c: No such file or directory
[t80040000] [t40008] me = 1 mytid = 262152
[t80040000] i_dump() 0 bytes total
[t80040000] pvmbailout(-11)

[t80040000] ready  3.3.0   Tue Jul 26 15:03:24 1994
[t80040000] removeshm: sem key = 40004: No such file or directory
[t80040000] removeshm: sem key = 40005: No such file or directory
[t80040000] removeshm: sem key = 40006: No such file or directory
[t80040000] [t4000b] libpvm [t
[t80040000] [t4000b] 4000b]: pvmbeatask() shmget: can't alloc msg buf: 
                             No space left on device

[t80040000] ready  3.3.0   Tue Jul 26 15:10:44 1994
[t80040000] removeshm: sem key = 40004: No such file or directory
[t80040000] removeshm: sem key = 40005: No such file or directory
[t80040000] removeshm: sem key = 40006: No such file or directory
[t80040000] removeshm: shm key = 4000b: No such file or directory
[t80040000] removeshm: shm key = 4000c: No such file or directory
[t80040000] removeshm: shm key = 4000d: No such file or directory
[t80040000] removeshm: shm key = 4000e: No such file or directory
[t80040000] removeshm: shm key = 4000f: No such file or directory
[t80040000] removeshm: shm key = 40010: No such file or directory
[t80040000] removeshm: shm key = 40011: No such file or directory
[t80040000] removeshm: shm key = 40012: No such file or directory
[t80040000] removeshm: shm key = 40013: No such file or directory
[t80040000] removeshm: shm key = 40014: No such file or directory
[t80040000] removeshm: shm key = 40015: No such file or directory
[t80040000] removeshm: shm key = 40016: No such file or directory
[t80040000] removeshm: shm key = 40017: No such file or directory
[t80040000] removeshm: shm key = 40018: No such file or directory
.
.
.
[t80040000] removeshm: shm key = 43b35: No such file or directory

*** logging truncated

Although I have tried to cleanup completely after each failure, there
seems to be something progressive about the problem. After the last 
failure, I was unable to restart PVM. The pvml.uid file from the restart
attempt follows.

[t80040000] mpp_init() can't create msg buffer: No space left on device
[t80040000] ready  3.3.0   Tue Jul 26 15:14:04 1994
[t80040000] i_dump() 0 bytes total
[t80040000] pvmbailout(-11)
[t80040000] mpp_cleanup() shmctl mybuf: Invalid argument


The fspmd example uses dynamic process groups, but I have had similar 
difficulty running a master-slave program which did not.

While it is clear that the problem arises at the operating system level,
I have no idea whether the fault lies with SGI's unix or with PVM. Many
thanks to anyone who can help!

Jim Tippett

