on 03-26-2014 12:54 AM
Hi everyone,
We are upgradeing from BW 7.01 to BW 7.4 and facing a very long runtime in the phase PARMVNT_SHD (actual phase PARMVNT_TRANS). It has been running for 3 days!
In this phase nametab activations are taking place. The upgrade is not stuck, but there are many objects to process from table DDXTT. I have checked the following:
1. We are using the latest SUM tool SUM 1.0 sp10
2. We are using the latest TP and R3trans
3. Our database is on Oracle 11g, latest patches applied and parameters are set based on SAP note recommendations
4. The hardware is HPUX with 8 cpus and plenty of memory.
It is just very odd that this phase takes so long and that there are so many objects to process.
Please help if you have any information or suggestions.
Many thanks!
Imran
Hi Imran,
We are facing similar issue, upgrading our ERP EHP5 (SPS10, NW702) to EHP7 (SPS03, NW741) using SUM 1.0 SP10 Patch 2. The ACT_TRANS phase took 19 hours. But the worst yet is the PARMVNT_TRANS phase which after 2 days only processed 94449 entries and still has 370967 in the DDXTT table to go. I too just logged a message and wondering if you ever got a resolution from SAP?
Elaine Wolf
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Elaine,
in my case the system was really slow and i doubled the db_cache shared_pool_size (plus reserve size ) and sga and restart the db
After this restart all worked fast as expected.
Maybe you check the Settings and try a restart of the DB
The sga was first set to 1 Gb and the system is running on linux so a small system..
I got in addition from sap
....
slow db access
....
1127194 - R3trans import with parallel processes;
1223360 - Composite SAP Note: Performance optimization during import;
744343 - Tips for importing Support Packages with minimized downtime;
558197 - upgrade hangs in PARCONV_UPG, XPRAS_UPG, SHADOW_IMPORT_UPG2
<(><-
589124 - Performance improvements when Support Package imported;
132536 - Improved semaphore mechanism in tp;
On oracle side:
354080 - Note collection for Oracle performance problems
618868 - FAQ: Oracle Performance
771929 - FAQ: Index fragmentation
588668 - FAQ: Database statistics
1171650 - Automated Oracle DB parameter check
983548 - Long runtimes during SAP Upgrade using Oracle 10
Please check the value of below sap note parameter and verify
1738989 - Upgrade/Support Package import hang due to INACTIVE wait
event "SQL*Net break/reset client" in Oracle
1635605 - CLIENT HANGS ON INSERT INTO TABLE WITH SECUREFILE LOB
...
Best regards
Thorsten
Hi Thorsten,
Thanks so much for your help. From your information and after analyzed our system that everything are in check except the parallelism, we zeroed in on note 1127194 and 1616401. We used the following commands to dynamically change the Parallel Processes:
1) cd <update directory>/abap/bin
2) SAPup set procpar gt=scroll to bump up all the parameters.
Then we terminated the tp process and also manually changed all the relevant profiles (the instance profile in both the regular and SUM directory and PARCONV.TPP) and then restart (ie Continue) from SUM - We noticed the performance improvement dramatically - It took 3 days to process 20% - And with the change it took 2-3 hours to complete the rest of the 80%. Now I just hope the rest will go smoothly.
Thanks again for sharing with much appreciation !
Best Regards,
Elaine Wolf
Hi Elaine,
how many process do you set up in the Setup in SUM and used at the end ?
Normally we use for d develop or test system , and this has no problem in these phase expect our small D System where the DB was the problem.
Batch Processes Uptime : 8 -> 5
Batch Downtime : 12 -> 6
SQL Process uptime : 8
SQL Process downtime : 12
R3trans uptime :8 -> 4 on P use 8
R3trans downtime : 12 -> 6 on P use 12
R3load process uptime : 8 -> 4
R3load process Downtime 12 -> 6
parallel phases Uptime : 6 -> 3
parallel phases Downtime : 6-> 3 on P use 6
Best regards
Thorsten
Hi Thorsten,
We had exactly the same issue with our dev-system, it is very small.
It was already running for more than 72 hours and it seemed the do nothing. The log files at ../SUM/abap/tmp/ where updated once in a while but not to much action.
After setting the new parameters and restarting the DB it is running rally fast.
Apart from yours we changed two more.
This is the list of parameters we changed in our system (we will be setting these back as soon as the upgrade is completed)
I'll give you all an update when this is completed.
Thanks to you all.
Hi all,
we've also such problem in the phase MAIN_TRANSEXEC/PARMVNT_TRANS. Our Upgrade is for a NW Gateway System 7.02 to 7.4 and there are some changes in the tmp and it seems that there are no double entries. The sapsr3.DDXTT changed also but not as i expected and from our experience this might be related to the new sum sp10 . With sp 9 P x we don't see such slow process.
I also open a sap call as i don't know how to speed up. The Files in the log directory don't changed since the Phase start only in the tmp and in the db, i see also two tp processes but the usage of ressources is not really high.
In the Shadow of our big systems the activation takes also a long time, but not ´like this performance,
The info from sap was you've many patches please wait it takes time...
In our case we decided for such small system use single so the system is down and when i callculate the entries in DDXTT this will take a long time for days !
Best regards
Thorsten Stracke
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi all,
well DMO is for the Migration to hana more or less...
We execute last time a sps update and there was this phase also in ~ 25 min ready.
This was on Solaris, now we run on Linux, but this shouldn't be a problem.
As the running system is a little bit small in case of memory i set 2 gb for the db from 1 gb and doubel also the shared plus db_cache_size as we got also alarms for the system.
The whole time beofre are 2 process for tp initiated by sum and one is del and one is activating nametabs.
After restart the db and restart the sum the del was running a short time and then finished. now the activation is in place and the tmp logs are updated in a short interval.
Maybe it was luck or the change of DB Parameters or the restart of the db...
Lets see how this go on.
Best regards
Thorsten Stacke
Hi Benjamin,
This is the top of the latest log which is in the SUM/abap/tmp directory. Its looking in DDXTT for entries to activate. As it activates it posts to this log. This particular log has been active for 24 hours (continuously updated). When I cancel the process thats when the logs are updated in the log directory. ALOG and SLOG.
There is a tp process running, and there is quite a bit of CPU load about 80% CPU utilization accross 8 cpus. In the past 18 hours it has activated 5,000 entries. It says it will only activate the first 7,000 tables but so far it has done over 100,000 entries.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
It is difficult to assist you if there are no relevant logs supplied.
Provide the latest logs from the <DIR_PUT>/abap/log and <DIR_PUT>/abap/tmp directories.
Regards
RB
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi everyone,
Thank you all for taking the time to reply, its very much appreciated. Please see my reponses below.
@ Gaurav Rana - resolving the nametab tables is true and critical for a Unicode Conversion before the export. But I am doing a release upgrade only. Before the upgrade these tables were empty, the entries are brought in by the upgrade.
@ Nicholas Chang - thank you for sharing these notes. I am not at that phase yet which is the XPRAS. Hopefully no performance issues will be faced there.
@ Sri Krishna Kanna - in the process of tuning this process I have stopped and repeated the phase from the point it left off many times, but it continues to want to process all entries from DDXTT. Started at 177,000 now at 68,000.
@ M. Abdul Jamil - this note is also for a different phase. Thank you for sharing.
I have a message open with SAP with High priority, its been about 2 days. After a follow up they said they will assign a processor to the message today. I will post back a solution, hopefully if one is found. I have reviewed all of the performance notes for activations during upgrades and especially the ones for PARMVNT_SHD but nothing applies.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
hi Imran,
Please go through this Note. I think, Issue will be resolve with the help of this note.
589296 - Problems w/ database links during the System Switch Upgrade
Regards,
majamil
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi,
Although it might not relevant to your case and problem, but worth checking below notes for performance improvement if you haven't go thru yet.
1629923 - Skip BW technical content objects activation during upgrade
1649901 - Time-critical processes in BW upgrade/Support Package
Thanks,Nicholas
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Imran,
There should not be any invalid objects present under tables DDXTT & DDXTF in your system before performing upgrade as per CUUC guide ,So here i suggest you to clear/delete entries from table DDXTT and re execute the phase.
Regards,
Gaurav
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
User | Count |
---|---|
86 | |
10 | |
10 | |
9 | |
6 | |
6 | |
6 | |
5 | |
4 | |
3 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.