AP Server sometimes hang when CPU load 100%

[former_member211576]

Hi experts,

Our system has migrated to ECC 6 and sometimes I find it slow when I

start a new transaction code. At that time, there are many work

processes idle but I find CPU load is up to 100%. I use

es/implementation = flat and es/use_mprotect = flase. Does this matter?

What is the root cause becuase I find just 4-5 background jobs running

will cause servers CPU load up to 100%. I use Dell PowerEdge 1955 with

2 dual cores CPU, 2.0GHz. I total have 9 AP Servers

[Former Member · Accepted solution]

HI,

You need to investigate more as far as CPU load is concerned. You will get better details in the OS06 or st06, in that can see which process is taking much amount of cpu time and then enact according to the same..

Rgds

Radhakrishna D S

[former_member211576]

I add more memory.

[former_member211576]

Hi experts,

thank for your reply. In fact, I have described in first post. Just 4-5 background jobs will consume all the CPU resources, each background jobs running consume one core with 100%.

[markus_doehr2]

There´s not much you (or we) can do about it.

You have some options though:

- if it´s a self written program contact the developers

- if it´s a standard program you could organize internally that those programs run at night and not during production daytime

- check with the users if their selection criterias are correct. Out of my experience I can tell that users are often too lazy to enter correct criterias.

Markus

[antonio_voce]

Describe the background process.

[Former Member]

>

> Hi experts,
>   thank for your reply. In fact, I have described in first post. Just 4-5 background jobs will consume all the CPU resources, each background jobs running consume one core with 100%.

That's not very unusual for background jobs. We have daily extracts here that write out GBs of data, each of my jobs take 100% of a CPU core as well. The way we get around it is to split the jobs up into different SAP application servers.

When the job is running, look into the process and see what it's doing. It should just be sitting there with the CPU usage count going up. This means that it's in a loop on a probably large internal table somewhere. You can try doing a SE30 trace on it, see which loop it is.

Additionally, it would also be wise to see if you can divide the data the job is processing into smaller pieces and run them sequentially.

Lastly, if all things above check out fine and you have no room for improvement, maybe it's time for a hardware purchase ..