OSI/MHS Configuration and Management Manual

Troubleshooting Your OSI/MHS Subsystem
OSI/MHS Configuration and Management Manual—424827-003
9-56
Problem: Insufficient Memory for Gateway
*GI-Backup-CPU.... -1
*GI-Priority...... -1
$posta subv 49> dsap $postc,short
Disk Space Analysis Program -- T9543D20 - (17JUN94) -- 12/2/94 12:39:25
Compaq Computers Incorporated 1981, 1983, 1985-1993
Free Space Short Report
-- Capacity (Mb) -- % -- Free Extents --
Volume (M) Total Free Free Count Biggest
$POSTC Y 895 6.31 0 591 4.20
The parameter values in the GPI_Initialize call are sufficient to permit several sessions
to be opened. Later, in processing an incoming message, the GIP process stops
running, and the following event messages appear in the event log:
94-12-09 14:53:10 \LAS.$PG11 TANDEM.MHS.D20 000021 $ZG1
GI.\LAS.GIGRP1 $PG11: Insufficient
Memory Space from swap file
\LAS.$POSTC.#2957
94-12-09 14:53:13 \LAS.$ZG1 TANDEM.MHS.D20 000018 \LAS.$ZG1:
Failure of process \LAS.$PG11 in group
GI.\LAS.GIGRP1 for reason 2
94-12-09 14:53:13 \LAS.$ZG1 TANDEM.MHS.D20 -00003
GI.\LAS.GIGRP1 changed from STARTED to
STARTING
94-12-09 14:54:26 \LAS.$ZG1 TANDEM.MHS.D20 -00003
GI.\LAS.GIGRP1 changed from STARTING to
STARTED
These messages indicate the GIP failure and the attempt by the MHS manager to
restart the GI group in recovery mode. (For detailed descriptions of these messages,
see the OSI/MHS Management Programming Manual). If the GIP fails again, the
MHSMGR does not restart the group: you must restart it explicitly using SCF.
When an incoming message causes the GIP to stop for lack of memory, that message
stays at the top of the queue. Until more memory is available for it, that message will
cause the problem to recur. To delete the message, change to a swap volume with
more free space on it, then have the application transfer the message in again and
delete it. (It might also be wise to examine the message to understand how the
problem occurred and how much space the message needed.)
To find out how much swap space a GPI application or the GIP is actually using, you
can use the TACL STATUS command, specifying the GPI application process or GIP
name, respectively. With the DETAIL option, you can find the names of the extended
segment, swap volume, and subvolume. Then you can use FUP INFO, DETAIL to find
out the size of the extended segment used for swap space. For example, if one GIP
can process a message that another GIP cannot process, you can find out how much
swap space the successful GIP is using and then make that amount available to the
unsuccessful GIP by changing the swap volume.