Performance Monitor question

[ca.portal.admin]

I cannot see any way to select by task number for the PMAM batch
reports. You can select by task code or task type, but not task number.
Does anyone know of a way to do this?

Kay Rozeboom
State of Iowa
Information Technology Enterprise
Department of Administrative Services
Telephone: [phone] Fax: [phone]
Email: [e-mail]

"
IDMS Public Discussion Forum
[e-mail]
SMTP
[e-mail]
[e-mail]
SMTP








Normal

Normal
"IDMS and TCP/IP, CA-Server 5.0 Issues and Notes"
"Hi gang. We've been experimenting with CA-Server 5.0 (JDBC type 4 connectivity) and the new TCP/IP interface. We've now got everything working nicely, but there were 5 issues we ran into which I thought I would pass along and perhaps save some of you some time. Note that these are not all in the same order in which they were encountered.

1.) On the client side, we were getting a ""JAVA.IO.UNSUPPORTEDENCODINGEXCEPTION:CP037"" because the JVM was installed without the ""support for additional languages"" option. Once we reinstalled the JVM with this option, the problem was verified resolved.

2.) The default for Windows TCP/IP connections (internal ports) is about 4000 (starts at port 1024, and runs up to 5000). The problem is that Windows doesn't free a socket for 4 minutes following the close, and it is easy to run out of ports. This results in the error message ""JAVA.NET.BINDEXCEPTION:ADDRESS ALREADY IN USE"" when attempting to connect. Microsoft has 2 tech bulletins regarding this problem. They are:

http://support.microsoft.com/default.aspx?scid=kb;en-us;196271

http://support.microsoft.com/kb/149532/EN-US/

I believe we chose the option to hack the registry, and the problem was verified resolved.

3.) We were running WebLogic 7.1 with no problems. When we attempting to start using WebLogic 8.1, we ran into a problem where we couldn't connect, and were receiving the error ""JAVA.LANG.CLASSCASTEXCEPTION"". We found that Weblogic 7.1 and 8.1 use proprietary connection pooling using the Driver Manager interface (WebLogic 9.0 uses the J2EE standard now). In WebLogic 7.1 there were no issues found during testing. In WebLogic 8.1, we got the exception. CA reported that they duplicated the problem and that it is a bug in WebLogic 8.1. The problem does not exist using WebLogic 9.0. BEA published a fix CR236961 to address this problem. We applied the patch, and the problem is now verified resolved. I have attached the patch to this e-mail.

4.) The sysgen compiler was doing strange things to my TCP/IP PTERMS. Things like not letting me delete them. CA put out a fix for this problem. It is QO71316. The problem desciption is as follows:
""SYSGEN ""DISPLAY ALL PTERMS"" displays duplicate PTERM entries for TCP/IP PTERMS sysgenned with
the 'PARM is' option on the PTERM statement."" I applied this APAR, and the problem is now verified resolved.

5.) Initially, I had difficulty understanding and setting up the listener on the TCP/IP line because of the ""STACK NAME'. I ended up using the IP STACK NAME IS '*DEFAULT' (note the asterisk). This will use the stack name of the host TCPIP, whatever it is. This works fine, and is nice because if we start the CV on another LPAR, the stack name of the new LPAR is used, not hardcoded in the PTERM (requiring a SYSGEN change). As an additional note, I used the standard '99xx"" where 'xx' is the CV number for this listener, as the listener port. This works well (unless you have more than 99 CV's - LOL).

Hope this helps some of y'all!

Rikki Kirschner
[phone]

"
IDMS Public Discussion Forum
[e-mail]
SMTP
[e-mail]
[e-mail]
SMTP








Normal

Normal
Re: Performance Monitor question
"Kay,
Not directly with ""canned"" Perfmon reports, but you can use Culprit
selection to look for a specific task number in the TASTSKID field.

Scott J. Brady
Holland America Line

[ca.portal.admin]

I cannot see any way to select by task number for the PMAM batch
reports. You can select by task code or task type, but not task number.
Does anyone know of a way to do this?

Kay Rozeboom
State of Iowa
Information Technology Enterprise
Department of Administrative Services
Telephone: [phone] Fax: [phone]
Email: [e-mail]

"
IDMS Public Discussion Forum
[e-mail]
SMTP
[e-mail]
[e-mail]
SMTP








Normal

Normal
Performance Monitor question
"I cannot see any way to select by task number for the PMAM batch
reports. You can select by task code or task type, but not task number.
Does anyone know of a way to do this?

Kay Rozeboom
State of Iowa
Information Technology Enterprise
Department of Administrative Services
Telephone: [phone] Fax: [phone]
Email: [e-mail]

"
IDMS Public Discussion Forum
[e-mail]
SMTP
[e-mail]
[e-mail]
SMTP








Normal

Normal
Re: Deadlocks
"Hello Joe:

The answer is no a task does not abend immediately, IDMS allows a dead lock
to occur and the tasks will wait on each other. Then the Dead Lock detection
mechanism (RHDCDEAD) wakes up at the sysgen defined Dead Lock detection
Interval looks at all dead locked takes and determines who to Kill (Dead Lock
Victims in the DMCT DIS STAT SYS display).

You can also issue a DCMT command to display and alter the dead lock
detection interval.

Starting with release 15.0 there are some new SYSIDMS Parameters, such as
ABEND_ON_DEADLOCK and DEADLOCK_DETAILS=ON, these were very helpful in debugging
my release 16.0 dead lock problems.

I use the IBM Book Manager Library for Windows and I have not had any
problems with it?

Bill Allen

In a message dated 10/21/2005 2:57:32 P.M. Eastern Daylight Time,
[e-mail] writes:

Hi All,
I have two questions....

1. When a task encounters a deadlock condition, does it ABEND immediately or
does it use one of the system timers?

2. Is there a place which explains how to read the IDMS documentation CDs
(I'm trying 15.0 right now)? I am trying Library Reader, Softcopy Reader
and Adobe and none of them are working real well. I know I had it figured
out at one time, but that knowledge has escaped through the various holes in
my head.

Thanks.

Joe Lupico
IDMS System Support
""Our World is a Happy World""

"
IDMS Public Discussion Forum
[e-mail]
SMTP
[e-mail]
[e-mail]
SMTP








Normal

Normal
Re: Deadlocks
"Joe,



To add to Chris and Bill's responses, you should know that the task does not
have to ABEND on a deadlock. It is true that once the deadlock detection
interval expires, the deadlock will be detected and one of the run units
will be ‘victimized’. The victim run unit is terminated but not the user’s
task. All the victim’s resources are released. The task that owns that run
unit is notified via IDMS-STATUS. At that point the RU is dead and no
successful calls can be made to it. However the task is free to continue.



Many programmers end the program unpleasantly when the deadlock status is
returned so it looks like IDMS forced the ABEND. This is neither necessary
nor desirable. I believe that it is better to reacquire/reinitialize
resources and retry the transaction from the top rather than blow up in the
user’s face. You can’t deadlock again on the same resource with the same
partner because the other task was allowed to acquire its desired resource
and has likely moved on. If it is still hanging around, like a batch job,
you will wait, but not deadlock. This is due to the fact that when you get
back to the point where you attempt to reacquire the resource you once had,
and that the other task wanted, he will have it and you will have to wait
for it.



When you successfully retry deadlocks this way, the end user doesn’t see the
deadlock though their response time will be higher; twice do the work twice
plus, on average, ½ the deadlock detection interval. It isn’t pretty but
it’s better than crashing.



Think of it this way…ABEND ½ of 1% of your transaction and your application
may become the butt of jokes at the water cooler. Retry them quietly and no
one will notice; which is a good thing.



Obviously it is best to eliminate deadlocks rather than deal with them.
YES, THEY CAN BE ELIMINATED. There are many techniques available but this
is not the place to explore them all. With that said, one of the best
things you can do is to acquire hot spot resources in the same order in
every program.



HAVE AN ABEND FREE DAY SOON!



Tom



PS: I solved my IBM Reader issues by using the PDF documentation.

[ca.portal.admin]

Does anyone know how much overhead is generated by collecting task-wait
statistics (""TSKWAIT=YES"" in #PMOPT)? The overhead I am interested in
is CPU.

Kay Rozeboom
State of Iowa
Information Technology Enterprise
Department of Administrative Services
Telephone: [phone] Fax: [phone]
Email: [e-mail]
"
IDMS Public Discussion Forum
[e-mail]
SMTP
[e-mail]
[e-mail]
SMTP








Normal

Normal
CA Education Survey
"As a public service we now post this survey link for CA.....

CA wants to hear your ideas in order to better align CA IDMS
training with your requirements and help focus CA's resources
to deliver the most effective CA IDMS training. Please take a few
minutes to complete the CA IDMS training survey at:
http://www.ca.com/us/products/collateral.aspx?cid=145946

(if you have any trouble with this hyperlink it is also in the NEWS
section of the IUA web page www.iuassn.org)

Thanks,
Terry Schwartz
"
IDMS Public Discussion Forum
[e-mail]
SMTP
[e-mail]
[e-mail]
SMTP








Normal

Normal
Re: Performance Monitor question
"I experimented with turning off PMIM serveral years ago and
could measure no appreciable different in cpu usage.

Thanks,
Terry Schwartz
perotsystems(tm)
[e-mail]
972-577-3722

[ca.portal.admin]

Does anyone know how much overhead is generated by collecting task-wait
statistics (""TSKWAIT=YES"" in #PMOPT)? The overhead I am interested in
is CPU.

Kay Rozeboom
State of Iowa
Information Technology Enterprise
Department of Administrative Services
Telephone: [phone] Fax: [phone]
Email: [e-mail]
"
IDMS Public Discussion Forum
[e-mail]
SMTP
[e-mail]
[e-mail]
SMTP








Normal

Normal
Re: [IDMSVENDOR-L] Performance Monitor question
"Hmmm. That might work for the applications that have steady use. (Some
of them vary a lot from day to day.) I don't know though - we have some
tricky users - we might have to keep the billing benchmark days secret!
:-)

[ca.portal.admin]

Does anyone know how much overhead is generated by collecting task-wait
statistics (""TSKWAIT=YES"" in #PMOPT)? The overhead I am interested in
is CPU.

Kay Rozeboom
State of Iowa
Information Technology Enterprise
Department of Administrative Services
Telephone: [phone] Fax: [phone]
Email: [e-mail]
"
IDMS Public Discussion Forum
[e-mail]
SMTP
[e-mail]
[e-mail]
SMTP








Normal

Normal
Re: Performance Monitor question
"I concur, since PMIM only takes a snapshot at set intervals, CPU is
negligible. PMAM is the costly one!

Regards,
Scott Brady

[ca.portal.admin]

Does anyone know how much overhead is generated by collecting task-wait
statistics (""TSKWAIT=YES"" in #PMOPT)? The overhead I am interested in
is CPU.

Kay Rozeboom
State of Iowa
Information Technology Enterprise
Department of Administrative Services
Telephone: [phone] Fax: [phone]
Email: [e-mail]
"
IDMS Public Discussion Forum
[e-mail]
SMTP
[e-mail]
[e-mail]
SMTP








Normal

Normal
Re: Performance Monitor question
"Kay

If not already mentioned, you also may wish check if task code USFAOPT
(Task Analyzer) is turned on.

On many sites when questioning its usage, application/dba teams replied.
....whats that?

Hence it got turned off and and teams informed how to start , stop and
use.....





Rob Klan/Cincinnati/IBM
Phone: [phone] (T/L: 349-2446)
ITN: 23492446
Email: [e-mail]



""Rozeboom, Kay [DAS]"" <[e-mail]>
Sent by: IDMS Public Discussion Forum <[e-mail]>
06/25/2007 10:46 AM
Please respond to
IDMS Public Discussion Forum <[e-mail]>


To
[e-mail]
cc

Subject
Re: Performance Monitor question






I don't think we can turn off PMAM because we use the stats so much.
But I am considering turning off PMIM. Do you know how much that might
save us?