HIGH DISC USAGE BY MANAGER.SYS IN GLANCE (426 Views)
Reply
Trusted Contributor
Scott Hower
Posts: 127
Registered: ‎05-28-2001
Message 1 of 9 (426 Views)
Accepted Solution

HIGH DISC USAGE BY MANAGER.SYS IN GLANCE

I noticed high disc usage by manager.sys on pin 55 yesterday and still have the same issue. System performance has been impacted only slightly but I would like to correct any issues that the system may have. The system is a 997-200 with 6.5 PPIII. The production volume set is a mirrored disk array in a Jamaica box (none of the mirrors are broken). The top lines from Glance:

P2 SYS MANAGER.SYS 2 LOAD B142 1.7% 0.7 0 0.0 MISC
P3 SYS MANAGER.SYS 3 B100 1.8% 0.0 0 0.0 MISC
P55 SYS MANAGER.SYS 55 A 13 3.7% 40.4 0 0.0 MEM
S0 0 MANAGER.SYS 193 RASERVER B149 1.4% 0.8 0 0.0 TERM
S0 0 MANAGER.SYS 200 RASERVER B149 1.9% 0.8 0 0.0 TERM
P554 SYS MANAGER.SYS 554 diaglogd C 13 8.2% 3.8 0 0.0 TERM

What could be causing this? What action can I take to correct it and prevent it in the future? Thanks!
Trusted Contributor
Scott Hower
Posts: 127
Registered: ‎05-28-2001
Message 2 of 9 (426 Views)

Re: HIGH DISC USAGE BY MANAGER.SYS IN GLANCE

I should have included a SHOWPROC...here it is:

:showproc pin=55;system
QPRI CPUTIME STATE JOBNUM PIN (PROGRAM) STEP

A13 * 5:18.217 WAIT 55 (..)
:

One concern I have is the priority that this process has - A13.
Valued Contributor
Mark Landin
Posts: 273
Registered: ‎11-13-1996
Message 3 of 9 (426 Views)

Re: HIGH DISC USAGE BY MANAGER.SYS IN GLANCE

Well the A queue is reserved for fundamental OS operations ... like paging to disk. Since P55 runs in that queue, and it's waiting on MEM, I'm wondering if it's not the MPE disk paging process. Heavy paging would, of course, induce heavy disk load. And this *would* be a performance hit!

Check your memory usage stats and your paging stats.
Trusted Contributor
Denis St-Amand
Posts: 89
Registered: ‎02-27-1997
Message 4 of 9 (426 Views)

Re: HIGH DISC USAGE BY MANAGER.SYS IN GLANCE

Did you verify which files that process was accessing in Glance ? Also, what is process name in Glance process screen ?

There has been an issue with DTS subsystem causing excessice I/O, that problem being solved in recent DTS patch.

There also has been issues when cstm log files became big and then have a huge number of extents, significantly slowing system down.

Under both /var/stm/logs/sys/ and /var/stm/logs/os directories, there are some files named log@.cur and diaglogd_activity_log. If one or both of those files have thousands of extents (you can see using a "listfile ,3"), then deleting it may both recover significant disk space and resume more normal performance.

Search ITRC MPE knowledgebase with "CSTM PERFORMANCE" to find out more details.
Trusted Contributor
Fred Metcalf
Posts: 233
Registered: ‎12-02-1996
Message 5 of 9 (426 Views)

Re: HIGH DISC USAGE BY MANAGER.SYS IN GLANCE

This could be a transceiver on a DTC that has the SQE (heartbeat) switch in the OFF position. By default it is in OFF position, while it should be ON when hooked to HP DTC. When it is OFF it causes a huge amount of heartbeat losses being logged to HP 3000 host.

Thank you to Martti of the HP MPE/iX Competency Center, North Europe, who helped me with this before.
Trusted Contributor
Scott Hower
Posts: 127
Registered: ‎05-28-2001
Message 6 of 9 (426 Views)

Re: HIGH DISC USAGE BY MANAGER.SYS IN GLANCE

Mark - the WAIT message for the process in GLANCE typically oscillates between MSG and MEM.

Denis - I recently cleaned those logs due to a lack of discspace and contiguous free sectors on my system volume set.
:LISTFILE /var/stm/logs/sys/@,2
PATH= /var/stm/logs/sys/

CODE ------------LOGICAL RECORD----------- ----SPACE---- FILENAME
SIZE TYP EOF LIMIT R/B SECTORS #X MX

1B BA 886168 2147483647 1 3472 * * activity_log
1B BA 2023672 2147483647 1 7920 * * diaglogd_activity_log
1B BA 263076 2147483647 1 1040 65 * memlog_activity_log
1B BA 80032 2147483647 1 320 20 * scan_hw_log


:LISTFILE /var/stm/logs/os,2
PATH= /var/stm/logs/

CODE ------------LOGICAL RECORD----------- ----SPACE---- FILENAME
SIZE TYP EOF LIMIT R/B SECTORS #X MX

16W HBD 0 67107839 1 32 1 * os/


:

Fred and Denis - That does seem like a possibility...in fact I must humbly admit that I've been bit by that several times before and always forget about that road. Denis, in fact, took my most recent call on that. Denis, I recall doing a LISTF or a SHOWPROC or something that told us that that was the problem. Can you refresh my mind on that? I did a LINKCONTROL @;STATUS = ALL to check heartbeat losses...I have two links and one has no heartbeat losses and the other has 150.
Trusted Contributor
Scott Hower
Posts: 127
Registered: ‎05-28-2001
Message 7 of 9 (426 Views)

Re: HIGH DISC USAGE BY MANAGER.SYS IN GLANCE

The process associated with pin 55:
proc(io_mgr_process)
Trusted Contributor
Denis St-Amand
Posts: 89
Registered: ‎02-27-1997
Message 8 of 9 (426 Views)

Re: HIGH DISC USAGE BY MANAGER.SYS IN GLANCE

Scott,

Then, it's probably a recurrence of DTC having a transceiver with SQE switch NOT set to "ON" or "ENABLED".

In Glance, looks at which files are accessed by process. You should see two files which names looks like "HxxxxxxA" and "HxxxxxxB". The xxxxxx are last 6 characters of offending DTC MAC address.
Trusted Contributor
Scott Hower
Posts: 127
Registered: ‎05-28-2001
Message 9 of 9 (426 Views)

Re: HIGH DISC USAGE BY MANAGER.SYS IN GLANCE

You're right Denis...same DTC as always, too. It's on a counter where it is easily accessible by people. I switched the SQE switch to ON - problem solved. Thanks all for your time; my apologies for my absentmindedness - I'll check there first next time!
The opinions expressed above are the personal opinions of the authors, not of HP. By using this site, you accept the Terms of Use and Rules of Participation.