cancel
Showing results for 
Search instead for 
Did you mean: 

Checpoint not complete hangs whole R3 system

Former Member
0 Kudos

We have R3 (4.6C) system running in Oracle 10.2.0.2 database and on HP-UX 11.31 IA operating system.

We patched operating system (HP-UX) and after that system hangs every time when comes "checkpoint not complete" situation??? We have had those "checkpoint not complete" situations before sometimes when database is "heavy loaded" but that was not a problem , oracle goes on after some 10 seconds. Now Oracle hangs for hours or forever??

Sometimes it goes ahead affter about 1 hour hanging and there is in oracle alert log error message: WAITED TOO LONG FOR A ROW CACHE ENQUEUE LOCK!

Usually we must restart database, that is not goooood!!

There is 20 CPU:s on system and 4 DBWR prosesses, redolog files are 200MB ( 8 groups, we had before 4 groups).

Do somebody have any tips?

BR. Arvo

Accepted Solutions (0)

Answers (3)

Answers (3)

Former Member
0 Kudos

Helo Arvo,

If I am not mistaken , I thing earlier you might have 2 redo log Groups having 2 Redo member in each group. Now you had increased number of redo log member in each group to 4.

Now,If the database is has heavy load. Then do Two things

1. Increase size of redo log member to 400MB each.

2. increase number of DB_WRITER_PROCESS= 5

3. change Parameter CHECK_POINT_INTERVAL=0

Regards,

PRATAP

Former Member
0 Kudos

Hello Pratap,

Jes, redos are like you say.

1. Increase size of redo log member to 400MB each.

-> not necessary yet

2. increase number of DB_WRITER_PROCESS= 5

-> maybe?

3. change Parameter CHECK_POINT_INTERVAL=0

-> it is.

These are standard methods when you want avoid "checkpoint not complete", but the error is that Oracle cant go over this situation as normally?

BR.

Arvo

Former Member
0 Kudos

Hi Arvo,

Check [this|; thread if that helps your situation.

Cheers,

Nisch

Former Member
0 Kudos

Hi

I think the 'checkpoint not complete' is already covered. Here some more information on 'WAITED TOO LONG FOR A ROW CACHE ENQUEUE LOCK!':

- when this happens a trace file is normally generated, check in ../saptrace/background and ../saptrace/usertrace. This file should contain more detailed information.

- please monitor the system, when the hang occurs, is there still cpu activity / which processes are running? Is only DML hanging / can you still select on v$session? It should be possible to identify the process blocking and only kill this one, so you don't have to restart the whole database

- it is highly probable that this issue is a bug. So be sure to contact support, if not already done.

- if you have a metalink (Oracle support) account, then please check this document:

Doc ID: Note:278316.1 - Potential reasons for "WAITED TOO LONG FOR A ROW CACHE ENQUEUE LOCK! "

( if you don't have a metalink account, check SAP note 758563 - Oracle Metalink access for SAP customers )

- As you have patched the os, please double check the patches, if possible try to verify, your io / disk operations are not significantly slower, than before, especially if you changed versions on filesystem or volume manager layer. As HP-UX V3 is pretty new, you might just be the first one hitting the problem, please get in contact with your HP support.

Best regards

Michael

Former Member
0 Kudos

Thanks,Nisch,

But, No help there...

Thanks Michael,

Tracefile did not show more (already checked).

We have alredy opened tickets to SAP Support/HP Support/Oracle Support......, this is maybe bug ;-( , we will see...

If somebody have more tips, please tell....

regads,

Arvo

markus_doehr2
Active Contributor
0 Kudos

First I would check that you have all necessary interim patches for Oracle installed, see

Note 871096 - Oracle Database 10g: Patch sets/patches for 10.2.0

to diagnose what is causing your database to stop see

Note 619188 - FAQ: Oracle wait events

The four DBWR may either not be enough (for a 20 CPU system and your load) and/or the underlying I/O subsystem is not able to write the data quickly.

Also check note 79341 - Checkpoint not complete which describes how do debug the situaiton.

Markus

Former Member
0 Kudos

Thanks markus,

Patches for Oracle are OK, we have already checked those Notes.

We can't see any performance problem in oper.system or in Oracle, system is very quick.

Seconds "Checkpoint not complete" is not a problem, totally hang situation is.

BR.

Arvo