Page Server service terminating

[MagicGirl]

Hello,

I have been running CR8 reports via an HTML page for several weeks with no problems. All of a sudden, yesterday and today, the Seagate Page Server service terminated in the middle of someone accessing a report. I restarted it with no problems, but would like to find out why and prevent it.

I have looked at the Crystal Decisions knowledge base, and the one listing I found regarding an automated install occurring does not appear to be the problem. I do not have the page server logging on, but will start that tonight. Does anyone have any suggestions?

Susan

 

[rosemaryl]

Dear Susan,

Were there any messages related to the Crystal Page Server Service in the Event Viewer?

ro Rosemary Lieberman
rosemary@microflo.com,

http://www.microflo.com

Microflo provides expert consulting on MagicTSD and Crystal Reports.

 

[MagicGirl]

Hi, Rosemary,

Yes - the message yesterday said "The Seagate Page Server service terminated unexpectedly. It has done this 1 time(s). The following corrective action will be taken in 0 milliseconds: No action."

This morning, it said "it has done this 2 time(s)." Cute.

There are no other error messages that occurred around the same time for both instances. I checked with one of our engineers, and there had been no server changes during the last couple days that we could pin this on.

Susan

 

[rosemaryl]

Dear Susan,

I am looking into it. In the meantime, if you do the following it will turn on logging for the Page Service which might help to diagnose the problem the next time it occurs.

If the Seagate Page Server is being run as a Windows NT service. Assuming that Crystal Reports has been installed on the C: drive, the following registry key:

HKEY_LOCALMACHINE\SYSTEM\CurrentControlSet\Services\pageserver

should be modified to contain the value:

"C:\Program Files\Seagate Software\WCS\pageserver.exe" -cache -deletecache -service -loglevel 9

The web server computer must be rebooted for this modification to take effect.

A log file, PageServer_Trace.log, containing information on the Seagate Page Server's activity will now be generated in the following location:

C:\Program Files\Seagate Software\logging

Also, was there anything atypical about the report this person ran? Were there parameters?

Are you getting this error running the report through Magic or was the report run from an url?

The only other information I have been able to find out so far indicates that grouping on the report can sometimes affect the Page Server Service. Did the report the user run contain a lot of groups?

Let me know,

ro Rosemary Lieberman
rosemary@microflo.com,

http://www.microflo.com

Microflo provides expert consulting on MagicTSD and Crystal Reports.

 

[MagicGirl]

Hi, Ro,

I did get to turn on logging Tuesday night, and rebooted the server, and did not see the error Wednesday. I have been out of the office today, so don't know if it reoccurred.

It happens when running reports from a Web page. Interestingly, the same reports could be run from the Reports form in Magic with no problem.

They are the same reports we've been running for several weeks, they were not new; I believe they have 2 or 3 group levels. They have date and/or Department parameters. Since the person who uses this page runs the same reports every day, I would expect that if the report is actually causing the error, i would have seen it before and/or it would have happened again after rebooting. So, when I get back in the office, Iwill check the log; hope it was a fluke, but it would be nice to know what caused it!

Susan

 

[rosemaryl]

Dear Susan,

Ah, so they are running directly from a link on a page?

I thought I saw some info on database connectivity issues with that. Let me research and I will get back to you.

I am wondering on the user side is she/he running the report several times in sequence?

Hmmm, this is interesting. What version of Magic/db and Crystal?
ro Rosemary Lieberman
rosemary@microflo.com,

http://www.microflo.com

Microflo provides expert consulting on MagicTSD and Crystal Reports.

 

[MagicGirl]

Hi, Ro,

One of the reports is usually run a few times in succession. This has been done on a daily basis since July, and only now did we see this shutdown of the Page Server service. Very strange!

We are using Magic 7.2, Crystal 8.0, and I did a full install of Crystal to get the Web Services on the Magic application server. It's worked very well for me thus far, and we plan to offer access to reports for select clients in a "portal" fashion.

I looked at the logs today, and didn't see that the Page Server service terminated again, but found the following repeated over and over:

Info: OSCASimpleFactoryMgrThread LocateNamingService
Info: OSCASimpleFactoryMgrThread AddUpFactories
Info: SIHello: --> AddUp
Info: SIHello: now check the naming service
Info: SIHello: now check the available hostports
Info: SIHello: <-- AddUp
Info: Warning: could not add factories
Info: OSCASimpleFactoryMgrThread RemoveDownFactories
Info: --> RemoveDown
Info: <-- RemoveDown
Info: ReportSourceRepositoryCache_Impl::registerServer

I am attempting to look this up on the Web, will have to guess that it is not affecting report access as we need it thus far. Anything look familiar about the above?

Susan

 

[rosemaryl]

Hi Susan,

I didn't get the message before I left the office for the day. I am going to look into this too and see what I find out.

By the way, heads up! If you plan on upgrading to Magic 7.5 it ships with Crystal 8.5 which does away with the Web Component Services. Crystal has placed this functionality into Crystal Enterprise.

Unfortunately, NAI chose not to ship Crystal Enterprise with the Crystal Report 8.5 it bundles with Magic - so you are forewarned that if you are rolling a &quot;portal&quot; solution out to your &quot;customers&quot; this will be an issue.

There are, of course, some work-arounds.

1. Get Crystal to send you Crystal Enterprise. 1 Magic user that I know was successful in getting Crystal to ship the CD as a replacement CD for $35.00, others were told that they would have to order Crystal Reports 8.5 directly from Crystal Decisions and they would receive Crystal Enterprise with it for free. Then you would install it and use the &quot;unmanaged&quot; reports option. That simply means running from an url as you are now, without logging into Enterprise.

2. Since you have a licensed version of Crystal 8.0 and the Web Component Server will run report written 8.5 - simply install Crystal Reports 8.0 onto another server(not Magic App or DB server) and you are in business.

If you have questions about that - feel free to contact me.

regards,

ro
Rosemary Lieberman
rosemary@microflo.com,

http://www.microflo.com

Microflo provides expert consulting on MagicTSD and Crystal Reports.

 

[jscranton]

Hi Rosemary,

I've been reading this discussion with interest.

I successfully turned on logging for the Crystal page service by editing the registry per your instructions above.

However, the web reports are now running quite a bit slower. I notice that a lot is written to the log for each report that is run.

Do you happen to know of a way to have the service create an abbreviated log, i.e., one that might simply give the name of the report that is being run?

Thanks & regards,
James

 

[rosemaryl]

Dear Jscranton:

The answer is to change the value of the loglevel switch:

-loglevel 9

Here are valid values:

Valid values are: 0,1,2,3, and 9. 9 generates the most information.

Try it with different values until you get the result (or as close as possible) to what you want!

ro Rosemary Lieberman
rosemary@microflo.com,

http://www.microflo.com

Microflo provides expert consulting on MagicTSD and Crystal Reports.

 

[MagicGirl]

Hi, James and Rosemary,

I am so glad to hear that someone else has felt a slowdown since implementing the logging. My primary reports user has complained of this, although she didn't really feel that it had started directly after the logging started, so I hadn't associated the two.

I did a Google search, and was luckily able to find this info:

http://support.crystaldecisions.com/nl/cr/2001_june.asp

Note the last parameters (-loglevel 09) are optional. -loglevel causes a log of all messages to be created, by default in c:\program files\seagate software\logging\pageserver_trace.log. 09 indicates the largest amount of logging possible will be done. You can set this anywhere between 01 and 09 depending on how much information you want. Be aware that at 09 the log will grow quickly and you should keep an eye on it so it doesn't fill your hard drive.

I would believe that whether or not we use the single or double digit entry, it would work, since the previous paragraph provides the Registry entry with the 09 instead of 9. I agree, level 9 has way too much info to be useful; I went through 500+ pages encompassing a mere 10 minutes worth of activity, one report request.

I am so glad to have seen your post, James, and am planning to modify the loglevel parameter in the next couple days to try this. Please let us know if you have tried it, too!

Susan

 

[MagicGirl]

Hi, Ro,

We responded at the same time! Thanks for the info on the valid values!

Susan

 

[jscranton]

Hi Rosemary & Susan,

Thanks for getting back to me. Right on point. Interesting that you ran into the same thing, Susan.

I had the loglevel setting changed to 1, and the speed of the reports went back to normal. However, the log is no longer being generated. I'll try setting &quot;2&quot; tomorrow and see what happens. I'll let you know.

One thing that puzzles me. Even with the speed restored to normal, a few of the larger reports still take a lot longer to run via the Web than with Crystal directly. For example, one of my reports takes 5 mins. to run on Web vs. 1 min. in Crystal. I know that the ActiveX controls need to be generated, etc., on the Web, but that wouldn't seem to account for such a large time differential. Any ideas why this might be?

Regards,
James

 

[MagicGirl]

Hi, James,

Thanks for the tip on the loglevel 1, I was going to try that, but will try 2 first.

Haven't run into the same performance problems, noticeably, over the Web, but I know there are tips and tricks for optimizing performance of reports (e.g., grouping, indexing, record selection, etc.). I would have to look them up, but they are usually in the User Guide and also in any how-to books you can find at the bookstore. Perhaps the additional steps taken to have the page server call the report exaggerate the time for each step in the generation of the report.

There might also be some documentation available through the Crystal Decisions Web site on optimizing Web performance. I happened to find one called cr8_web_best_practices.pdf, which talks about the installation and configuration of the Web components server, and I know there are probably other documents available which give advice on report performance (maybe the Web Reporting Administrator's Guide). Did you happen to look through the log file at all when it was generated at level 9? It looks like there are numerous steps in the process of calling the report, and they take place continuously (it looked like once a second, and I am not kidding!!). It looks almost like some sort of cache refresh or something. So, probably the reports that are slow are hitting a step or two which take a long time during each one of those calls.

Thanks again for the tip - hope you find something which can help you, I'd love to hear what it was.

I haven't had a repeat of the service termination since I bounced my Magic/Crystal application server, so I am hoping that the specific issue which started this thread was just a fluke....

Susan

 

[jscranton]

Hi Susan,

Thanks for the info. It turns out that when I run the report from the machine that hosts the web server in CRYSTAL (not over the web), it still takes 5 mins.

When I run the same report from my machine in Crystal, it takes 1 min.

So for some reason, Crystal runs much slower on the server machine than it does on mine. It's not a web-related slowdown.

The slowdown lies in the reading of the records from the database. On the web server machine, the report reads about 400 records per second. On my machine, it reads about 2000 per second.

So the question I need to find out the answer to is, why would Crystal run so much slower on one machine than another.

I'll let you know if I find anything. Would love to hear about your loglevel experiments, etc.!

James
jscranton@hotmail.com

 

[MagicGirl]

Hello,

I have had the loglevel on the Page Server set at 2 for a couple weeks now. The report response time has greatly improved, but I see nothing in the log at all. Maybe this is a good thing? I am wondering if loglevel 2 is a good intermediate step, where it would show problems but not every cache refresh as 9 does? I would love to find out specifically what is expected to show at each level, so that I know whether or not it is too low.

Susan