Only 2 scheduled processes executing

[sethron]

Hi,

on CA 11.1.7, we have schedules that are run at 07:00 every morning. They are all fast reports that run 5-10 seconds each and send an email. In the previous 11.0 version, they would execute 16 at a time, since my batch number of processes was set to 16 (2 high, 4 low affinity connections, same parameters during peak and non peak and for reportservice. I've set it both on dispatcher level and individual). Same settings on 11.1.7. Query service JVM size is also at 100GB max.

But on the 11.1.7., they would only execute 2 at a time, while rest are in Waiting status, waiting to be executed.

Is there a new setting in 11.1.7?

[MFGF]

Hi,

on CA 11.1.7, we have schedules that are run at 07:00 every morning. They are all fast reports that run 5-10 seconds each and send an email. In the previous 11.0 version, they would execute 16 at a time, since my batch number of processes was set to 16 (2 high, 4 low affinity connections, same parameters during peak and non peak and for reportservice. I've set it both on dispatcher level and individual). Same settings on 11.1.7. Query service JVM size is also at 100GB max.

But on the 11.1.7., they would only execute 2 at a time, while rest are in Waiting status, waiting to be executed.

Is there a new setting in 11.1.7?

Hi,

In terms of the number of concurrent requests the system can handle, it works as follows:

- Each connection for a report service can handle 1 request. If you have only 1 report service configured and 4 connections, this means the system can run 1 x 4 = 4 requests concurrently.
- Each report service you define multiplies this. If you have 16 report services configured, each with 4 connections, this would give you 16 x 4 = 64 requests concurrently.

I would expect to see all your reports attempting to run concurrently with the way you have configured things. Are the reports using CQE or DQM? If they are DQM, then your Query Service memory setting should be more than enough. Does your system have enough memory to support 100Gb being allocated to the Query Service alone?

This one is a puzzle!

Cheers!

MF.

[sethron]

Thank you for the answer. All of the reports are DQM.

The server has 128GB RAM, so is 100GB for Query Service too much? What are some guidelines? Can i bump it up more just to be sure?

When you say "Each report service you define multiplies this.", do you mean the setting under ReportService (or equivalent Batchservice) "Maximum number of processes for the report service "

I remember reading an other post you made with the formula that goes something like "Maximum number of processes" x "Number of low affinity" + "Maximum number of processes" x "Number of high affinity" = Total number of requests concurrently. Did I understand that correctly?

[MFGF]

Thank you for the answer. All of the reports are DQM.

The server has 128GB RAM, so is 100GB for Query Service too much? What are some guidelines? Can i bump it up more just to be sure?

When you say "Each report service you define multiplies this.", do you mean the setting under ReportService (or equivalent Batchservice) "Maximum number of processes for the report service "

I remember reading an other post you made with the formula that goes something like "Maximum number of processes" x "Number of low affinity" + "Maximum number of processes" x "Number of high affinity" = Total number of requests concurrently. Did I understand that correctly?

Hi,

Yes, the number of processes for the report service, that's correct. Just sum up the number of connections (low and high) and multiply that by the number of processes for the report service - this will give you the maximum concurrent throughput. Eg 3 low affinity connections, 1 high affinity connection, 6 processes for the report service would mean (3 + 1) * 6 = 24 concurrent requests supported.

Cheers!

MF.

[sethron]

Great, thank you for the input.

Do you have any recommendations for Query Service heap size?

[MFGF]

Great, thank you for the input.

Do you have any recommendations for Query Service heap size?

I would set the initial heap size to 4Gb, and try with an 8Gb max size to begin with. If you're seeing issues, up the max a little - maybe to 12Gb.

Unless you're using Dynamic Cubes, this should be more than enough.

Cheers!

MF.

[sethron]

Isn't 12GB for Query Service a bit to low?
My current high watermark for QS is around 40GB, so I put the maximum heap size at 64GB (servers has 128GB of RAM, with 32GB dedicated to WebSphere, that leaves pottential 96GB for QS ).

[MFGF]

Isn't 12GB for Query Service a bit to low?
My current high watermark for QS is around 40GB, so I put the maximum heap size at 64GB (servers has 128GB of RAM, with 32GB dedicated to WebSphere, that leaves pottential 96GB for QS ).

Are you using Dynamic Cubes? I don't think I've ever seen the Query Service need anything like that unless you have large Dynamic Cubes in the mix. What kinds of reports are you running that require 12Gb of memory for rendering all at once?

Cheers!

MF.

[sethron]

We have several Dynamic cubes that are all together around 4GB in size. During peak hourse 8-10 around 200-300 hundred consume reports.

[sethron]

Hi,

on CA 11.1.7, we have schedules that are run at 07:00 every morning. They are all fast reports that run 5-10 seconds each and send an email. In the previous 11.0 version, they would execute 16 at a time, since my batch number of processes was set to 16 (2 high, 4 low affinity connections, same parameters during peak and non peak and for reportservice. I've set it both on dispatcher level and individual). Same settings on 11.1.7. Query service JVM size is also at 100GB max.

But on the 11.1.7., they would only execute 2 at a time, while rest are in Waiting status, waiting to be executed.

Is there a new setting in 11.1.7?

UPDATE: we found that it wasn't the batch service fault, they were indeed executing multiple reports, but were stuck in waiting while waiting for the emails to be sent (one report can send up to 30 email due to bursting).

The schedules and burst reports were created on a 11.0 version. A report has burst to users based on the company, but if there is no data for the company, the mail with XLS  attachment wasn't sent. My email as the schedule owner was nowhere in the burst list.
Now on the 11.1.7. FP4 IF8, I get an email when the schedule for the report is run and I also get all the 30+ emails with empty or non empty data.
If the burst output is empty, only I get the email with empty body and no attachment.
If the burst output is not empty, burst users and me get the email with attachment (my email is not in the burst key and it's always attached to the end)

I've tried the following:
-Deleting old schedule and creating a new one - no luck
-creating a new schedule, deleting my mail from the To: part and then activating Burst option in schedule - no luck, i get "CNC-SDS-0336 There are no valid recipients for the message. The message will not be sent. Please check delivery addresses." and then it sends 30+ emails with the same failed to send message
-creating a new schedule, then enabling burst options in schedules without deleting my mail - no luck
-running the same steps as above, but using the report option run with owner capabilities instead of with users credentials

Is there a fix for this?