Performance drop from JobServer vs. PageServer on Same RPT

[MJRBIM]

Crystal Enterprise 8.5 installed on a single-server, Reports use Oracle native drivers.

I have a published Report Object with a predefined Database Login/PW. It has 7 parameter prompts at runtime - the examples below are both run with the same parameters.

EXAMPLE "A" : When a user VIEWs the Report, the PageServer returns the 2-page report in about 30 seconds.

EXAMPLE "B" : When a user SCHEUDLEs the Report, the PageServer returns the 2-page report in about 12 minutes.

I am stumped as to why the two methods of running the SAME report, against the SAME datasource, with the SAME parameters would be so wildly different.

The PageServer seems to be running and configured fine, and we don't experience this with any other reports in the same environmnet.

 

[Turkbear]

Hi,
IIRC, the Job Server is in charge of Scheduled reports, so check its configuration - also how are the parameters passed when scheduled?

 

[MJRBIM]

Sorry, typo in my original message...should have said JobServer for the latter parts...

- - - - - - - - - - - - - - - - - - - - -

EXAMPLE "B" : When a user SCHEUDLEs the Report, the JobServer returns the 2-page report in about 12 minutes.

I am stumped as to why the two methods of running the SAME report, against the SAME datasource, with the SAME parameters would be so wildly different.

The PageServer and the JobServer both seem to be running and configured fine, and we don't experience this with any other reports in the same environmnet.

- - - - - - - - - - - - - - - - - - - - -

...could you clarify what you mean by "how the parameters passed when scheduled"?

They are paramters within the RPT file that are used as part of the Crystal SELECT statement.

Thanks in advance!

 

[Turkbear]

Hi,
I meant how, when run as a scheduled report, are the prompts handled?

Do they have fixed default values?

 

[MJRBIM]

No default values for parameters, all values are "EMPTY".

They are entered into the PARAMETER tab of the SCHEDULE window in ePortfolio when the job is scheduled.

 

[Turkbear]

Hi,
Thanks..I don't use ePortfolio much, so I forgot about that tab..

 

[rhinok]

What viewer are you using? ActiveX? The ActiveX viewer doesn't return the entire report, it only returns a page at a time, which is one reason it may appear to run faster. Also, the report may be cached, which would improve return time significantly. Lastly, Page Server jobs are run as threads, which require fewer resources than Scheduled Jobs.

Sheduled reports are fully formatted through the process and may take slightly longer in general. Also, if a report is exported to a different format or sent to a destination (such as FTP or SMTP)as part of the scheduling process, it may take longer still. These are all things that are handled by the job server, but would be done manually by the User for a report that is Viewed on demand. Lastly, Job Server jobs run as processes, not threads, so they require more resources to run.

Other considerations[ol][li]Are your Page Server and Job Server on the same machine?[/li][li]How many Page Servers and Job Servers do you have?[/li][li]How are the various servers configured?[/li][li]Are these times consistent or do they fluctuate? If they fluctuate, what is the load on the Report Server at schedule time - the report may just be waiting in a Pend status until the resources are freed up[/li][li]Scheduled report instances are saved in the File Repository Server - Is the FRS installed on the same machine as the Job Server?[/ol]As you can see, there are a number of issues that could cause this, but you need to troubleshoot them to find out what the root cause is.

 

[MJRBIM]

What viewer are you using? ActiveX? The ActiveX viewer doesn't return the entire report, it only returns a page at a time, which is one reason it may appear to run faster.

I am using the ActiveX viewer, but the Report is ONLY 2 pages long, and is delivered as a report (not a PDF, Excel, etc).

Are your Page Server and Job Server on the same machine?

YES, in this development environment all servers are on the on the same machine.

How many Page Servers and Job Servers do you have?

In this developemet environment there is one (1) of each.

How are the various servers configured?

Crystal Page Server
Location of Temp Files: C:\Program Files\Crystal Decisions\WCS\Data\shqeihcedev.pageserver
Maximum Simultaneous Processing Threads: 75
Minutes Before an Idle Job is Closed: 20

Crystal Job Server
Object Type: CrystalEnterprise.Report
DLL Name: procReport
Maximum Jobs Allowed: 5

Are these times consistent or do they fluctuate? If they fluctuate, what is the load on the Report Server at schedule time - the report may just be waiting in a Pend status until the resources are freed up.

These times appear constant. I only get the PENDING status in History of a few seconds, it shows RUNNING for the rest of the 12 minutes.

Scheduled report instances are saved in the File Repository Server - Is the FRS installed on the same machine as the Job Server?

YES, in this development environment all servers are on the on the same machine.

- - - - - - - - - - - - - - - -

The report that is acting "weird" uses the ORACLE native driver to access the database. Other reports that do not appear to have this problem use the Microsoft ODBC for Oracle driver...could that be part of the issue?

 

[rhinok]

Using the Oracle Native driver may cause the issue. Are there differences between the TSNAMES.ora file and the ODBC DSN you use?

In this case, it may be a good idea to test a version of the report that uses the Microsoft ODBC driver you're using for all other reports to see if this alleviates the problem.

 

[alb]

Remember the Pageserver will process only one page at time. The Jobserver process the whole report and saves the instance of the report in the output directory.

Pageserver threads really has nothing to do with the connection to the database.

I would not recommend using Microsoft ODBC driver for Oracle since it not support by Crystal Decisions. Use the Oracle ODBC driver instead. The Oracle native driver may or may not be a multithreaded driver because it depends on the version you are using. What version are you using? Also are the Oracle client the same version as the Oracle server?

 

[wilsonmichael]

Alb

i am not having the problem in his post but i am having thread locking issues and cant seem to find out why. All i know is the logon function to the APS sometimes does not respond and just sits there waiting never returning an error. I have created a thread called "Logon function to APS stops responding" You can look there for more detail if you need it.

My APS is pointed to an oracle instance and i am using an oracle native driver to connect to it. what versions are multi-threaded and how do you check what version of the oracle native driver you are using.

i have oracle 8i installed and use oracle provider for ole db 8.1.7.0.0 in my ASP code.

Thanks

Mike

 

[MJRBIM]

After much digging around, head scratching, and shaking of voodoo chicken bones....I found it!!!!

Someone had set a trace on the command line for the JobServer in the JobServer properties in the Crystal Configuration Manager.

It looked like this....

"\\SERVERNAME\C$\Program Files\Crystal Decisions\WCS\pageserver.exe" -service -name shqeihcedev.pageserver -ns shqeihcedev -restart -nocrheapalloc -trace

Trimmed out the trace, restarted the server and now performance is similar to the PageServer.

 

[Turkbear]

Hi,
Good catch..Now find out where the trace output was sent, since it may be a large file(s)..

Then find out whodunnit...