users said... the system is too slow..... and we type sm50?

Basis (Basis Technology Modules: Basis Component/System Administration, GUIs)

Moderators: Snowy, thx4allthefish

thx4allthefish
Posts: 5694
Joined: Sat Oct 26, 2002 6:18 pm
Location: barolo barrel

Post by thx4allthefish » Tue Oct 29, 2002 4:04 am

hey yls177

you're mixing things up here.

sara has nothing to do with online - redologs!! sara will of course archive your business data and compress them. they will be stored as flat files in the path you specify in the customizing.

online-redologs are a different topic. they are created by the database to enable crash recovery of the database in case of emergency.

go have a look around in http://www.service.sap.com/notes on the topic of redologs and sara. maybe this will clear the matter a bit

thx4allthefish
Posts: 5694
Joined: Sat Oct 26, 2002 6:18 pm
Location: barolo barrel

Post by thx4allthefish » Tue Oct 29, 2002 4:05 am

hey yls177

you're mixing things up here.

sara has nothing to do with online - redologs!! sara will of course archive your business data and compress them. they will be stored as flat files in the path you specify in the customizing.

online-redologs are a different topic. they are created by the database to enable crash recovery of the database in case of emergency.

go have a look around in http://www.service.sap.com/notes on the topic of redologs and sara. maybe this will clear the matter a bit

yls177
Posts: 4016
Joined: Thu Oct 24, 2002 9:12 pm

Post by yls177 » Fri Nov 01, 2002 4:13 am

1) in sm50, if the work process is the type background, and is in office working hours, the understanding is that it should not run and snatch precious resources. am i right?

2) sm66 is for the systemwide work process. so given a sid, there will be only one instance or the most a logon group (which still is one instance). Therefore, what difference does sm50 and sm66 had?

thanks

bubu
Posts: 529
Joined: Mon Oct 21, 2002 7:42 am

Post by bubu » Fri Nov 01, 2002 9:47 am

Hi yls177,
1) in sm50, if the work process is the type background, and is in office working hours, the understanding is that it should not run and snatch precious resources. am i right?
you are right but remember that there are jobs run by applications (managers) that need to be executed during office working hours ... the idea is to minimize if not possible to completely eliminate them ...
2) sm66 is for the systemwide work process. so given a sid, there will be only one instance or the most a logon group (which still is one instance). Therefore, what difference does sm50 and sm66 had?
sm66 checks for internal RFC processes

braveheart

Post by braveheart » Fri Nov 01, 2002 10:25 am

The offline redo logs in saparch are not comressed. You can use brarchive to compress them to another area of disk. Set up a regular job in DB13.(brrestore will uncompress them)

Guest

Post by Guest » Tue Nov 05, 2002 12:29 am

sm66 checks for internal RFC errors.

does that mean between a central instance and its application servers?

thanks

yls177
Posts: 4016
Joined: Thu Oct 24, 2002 9:12 pm

bubu, can clarify..

Post by yls177 » Thu Nov 07, 2002 10:03 pm

what do u mean by internal RFC errors...

yls177
Posts: 4016
Joined: Thu Oct 24, 2002 9:12 pm

Post by yls177 » Thu Jul 17, 2003 9:06 pm

Anonymous wrote:sm66 checks for internal RFC errors.

does that mean between a central instance and its application servers?

thanks
i think u are right...

but can someone confirms?...

thanks

bubu
Posts: 529
Joined: Mon Oct 21, 2002 7:42 am

Post by bubu » Fri Jul 18, 2003 6:48 am

sm66 checks for internal RFC errors.

does that mean between a central instance and its application servers?
Its internal RFC processes (not errors). And its not only between CI and its application server, but also between the different work processes. Normally what you see there are processes communicating with the DB ('cause these are one of those processes which take time to complete)

Regards,

MBirch
Posts: 387
Joined: Thu Nov 14, 2002 8:15 am
Location: Leeds UK
Contact:

Post by MBirch » Fri Jul 18, 2003 12:49 pm

SM66 is useful to check ALL active workprocesses in a multi server system at the same time. It's my home screen.

I've had many and various performance problems over the past two years - obscure things to check that haven't been mentioned are:

Disk queue length - we had crap old disk array - and many subtle problems stemmed from really crap disk performance.

Network speed - When ping is simply not enough - use NIPING delivered by SAP - it measures throughput between sap servers and works at socket level. We had a wierd NIC problem dropping to 10Mbit - only diagnosed this through niping. Server checked ok and ran fine but boy was it slow. Check OSS, niping does what it says on the packet (pardon the pun).

Zcode - Always get your abap guys to tune their code - most don't know about the performance analyser for abap - SE30 in our crap 3.1i version.
Should give them a lot of clues where their abap is bad.

I agree with all the other good advice - good luck!

MB

bubu
Posts: 529
Joined: Mon Oct 21, 2002 7:42 am

Post by bubu » Fri Jul 18, 2003 1:19 pm

SM66 is useful to check ALL active workprocesses
Doesn't show up all active processes ... as message number S1249 says:
".. RFC calls" ... only for internal RFCs ....

BUT transaction is very helpful in determining processes that taking too long to execute than necessary ...

MBirch
Posts: 387
Joined: Thu Nov 14, 2002 8:15 am
Location: Leeds UK
Contact:

Post by MBirch » Fri Jul 18, 2003 2:57 pm

Hi Bubu,

you can change the settings to view more but as a snapshot of what's happening its the most informative screen.

Did the original performance problem get solved?

I think that unless SAP points to a specific problem without digging too hard then you've gotta look at OS and server configuration as well for your performance problems.

Oh yeh, remembered another perf problem:

We had really slow disk access as I said.
This caused unexplained exclusive locks at db level leading to blocking of workprocesses and the system grinding to a halt.

In MSSQL7.0 if a stored procedure (ABAP code compiles to stored procedures on MSSQL systems) took longer than 1000ms to return it's data then the system would re-compile the stored procedure. The next run should be much quicker. The thinking behind this is that not all selects are efficient and that a change of selection at user level might benefit from a recompile. Our disks were so slow that we bust this limit very often, and even got stuck in a recompilation loop. And into the bargain no other process can use this abap code while the blocking process is recompiling - exclusive lock waits went through the roof and I had to kill healthy processes just because our disks were slow. Aaarrggh!

This one took months to resolve (increased the rsdb/mssql/max_duration
1000ms to 5000ms only helped reduce) and eventually upgraded H/W.

I mention the story to illustrate that there isn't always a quick fix and that there can be a loooong chain of events between root cause and symptom.

Good luck with those complaining users! Be tenacious - it can take time.

MB

oldsapguru
Posts: 5489
Joined: Wed Oct 23, 2002 6:08 pm
Location: Nashville via Manhattan
Contact:

Post by oldsapguru » Mon Jul 21, 2003 10:43 pm

1. What has changed since the performance is poor?

System Parameter Changes (RZ10)?
Increase in the number of users?
Kernel patch?
Support Packages?
OS patches or PTFs?
Hardware Changes?

2. Any signs of general problems in the System Log - SM21?

3. Places to look:

SM50: -> Process -> Trace -> Components – is there a Trace level set? Is it greater than 1? What is the utilization of the work processes? Click the white clock picture-icon. Is the total CPU for the last dialog process > 10 minutes?
ST04: Is the database monitor activated? If yes, it should deactivated it in normal operation. ST02: Check the buffering quality. If paging occurs in a buffer, the corresponding parameter should be increased. Also refer to SAP Note 121625 in this context.
ST03: -> Select a server -> Today's Workload: What are the response times? In the case of poor response times, where is most of the time needed? Button 'Top Time': Are there a lot of different transactions with poor response times or are there only a few? Is a certain transaction always slow or only sometimes?
ST06: -> Detail analysis menu -> Hardware Info. To which extent are the hardware resources utilized? Call ST06 at times with poor system performance. What is the CPU utilization? -> Goto -> Current Data -> Snapshot -> Top CPU processes and display the main CPU consumers. What is the utilization of the disks? How high is the paging in the base pool?

4. If only few transactions are affected by the performance problem, you should additionally note the following points:

1. Check whether modifications were made in the affected or related transactions.
2. ST03: Display the corresponding performance records.
3. SE30: Carry out a runtime analysis for the affected transactions.
4. ST05: If the database times are high, you should generate an SQL trace of the affected transactions to find out whether the database access is carried out in a useful way.

5. Look for tables that have experienced unusual monthly growth.

1. DB02 -> Space Statistics button
2. Press Enter on the Tables and Indexes popup
3. History -> All objects off/on
4. Click the Months button
5. Click on the first number under the Rows – Chg/Month header and click the Sort button.
6. The sorted results shows the top tables when it comes to rows changes per month. These tables are your “database hogsâ€

Locked