_mprosrv.exe processes running on the server - 100% used each morning

[Kenmath]

Hi,
I'm facing to a strange pb since many years.
It's very difficult to login to Syte all morning when people are coming.
During 10-15 minutes some _mprosrv.exe processes are using all procesqsor ressources on the Syte server.
During this time you see the Syte client starting screen but it's frozen before the login screen.
After many minutes all is fine without doing anything... strange...

We are starting the client with a .bat file :
start /i S:\DLC83C\bin\prowin32.exe -basekey INI -ininame S:\reel\SymixClt-Local.ini -pf S:\Reel\pp39run.pf -p menu/symix.p -T %TEMP%

All days, same situation same time...when 5-6 users trying to login Syteline...

Does anyone was already facing the same pb ?

Server:
Windows NT 4
Progress 8.3C db
SyteLine v4.02.10D

Thx
BR
Mathieu

 

[Cringer]

I'm going to move this to the dba thread as I think it's more suited for that. I have a couple of initial questions:

1) What's the contents of the pf file?
2) Can you post the first chunk of lines from your database log since the last restart of the db. The first 70 or so lines from a db start should do. That way we (Tom ) can see what the db configuration is and see if there's anything that might be obvious.

 

[Kenmath]

Hi Cringer,
Thx for your answer.

Here the content of my pf file :

# SESSION,Session-related parameters
-Bt 500
-D 350
-TB 4
-TM 15
-h 12
-inp 5000
-mmax 32768
-nb 255
-q
-rand 2
-yy 1950
-d dmy
-E
#
# MAIN,pp39
-db pp39
-B 5000
-N tcp
-S pp39
-H zeus
#
#
# CFG,pp39cfg
#-db s:\pp39cfg
#-B 500
#
# GLOBAL,symglbl
-db symglbl
-ld symglbl
-B 500
-N tcp
-S symglbl
-H zeus
#
#
# LICENSE,symlic
-db symlic
-ld symlic
-B 500
-N tcp
-S symlic
-H zeus
#
# EXTFIN,ppextfin
-db pp39fin
-ld extfin
-B 500
-N tcp
-S ppextfin
-H zeus
#
# EFAS, EFASlive
-db efaslive
-ld efas
-B 500
-N tcp
-S efaslive
-H zeus
#

Answer of the 2nd answer in a next quick reply in a few minutes

 

[Kenmath]

Here pp39.lg after server restart monday 8 PM :

Mon Apr 23 20:26:58 2012
20:26:58 BROKER 0: Début de la session Multi-Utilisateurs. (333)
20:26:58 BROKER 0: A démarré pour pp39 en utilisant tcp, pid 246. (5644)
20:26:58 BROKER 0: PROGRESS Version 8.3C sur WINNT. (4234)
20:26:58 BROKER 0: Serveur lancé par Administrator sur CON:. (4281)
20:26:58 BROKER 0: Nom Physique de la Base de Données (-db): d:\sytedb\reel\pp39. (4235)
20:26:58 BROKER 0: Database Blocksize (-blocksize): 8192. (6573)
20:26:58 BROKER 0: Type de la Base de Données (-dt): PROGRESS. (4236)
20:26:58 BROKER 0: E/S Direct (-directio): Not Enabled. (4238)
20:26:58 BROKER 0: Nombre de Buffers pour la base de données (-B): 34000. (4239)
20:26:58 BROKER 0: Excès de la taille de la Mémoire Partagée (-Mxs): 16419. (4240)
20:26:58 BROKER 0: Taille actuelle de la 'Lock Table' (-L): 2016. (4241)
20:26:58 BROKER 0: Entrée de la Hash Table (-hash): 9337. (4242)
20:26:58 BROKER 0: Valeur actuelle de 'Spin Lock Tries' (-spin): 0. (4243)
20:26:58 BROKER 0: Le nombre de sémaphores (-semsets) 1. (6526)
20:26:58 BROKER 0: Crash Recovery (-i): Enabled. (4244)
20:26:58 BROKER 0: Retard du "Before-Image Flush" (-Mf): 3. (4245)
20:26:58 BROKER 0: Nom du Fichier Before-Image (-g): (4246)
20:26:58 BROKER 0: Fichier E/S Before-Image (-r -R): Reliable. (4247)
20:26:58 BROKER 0: Intervale du truncate du Before-Image (-G): 60. (4249)
20:26:58 BROKER 0: Taille de Cluster du Before-Image: 512. (4250)
20:26:58 BROKER 0: Taille du Bloc Before-Image: 8192. (4251)
20:26:58 BROKER 0: La taille du fichier BI Threshold (-bithold): 0. (6550)
20:26:58 BROKER 0: Le fichier BI Threshold Stall (-bistall): Désactivé. (6552)
20:26:58 BROKER 0: Nombre de Buffers du Before-Image (-bibufs): 5. (4252)
20:26:58 BROKER 0: Nom du Fichier After-Image (-a): Not Enabled. (4253)
20:26:58 BROKER 0: After-Image Stall (-aistall): Not Enabled. (4254)
20:26:58 BROKER 0: Taille du Bloc de l'After-Image: 8192. (4255)
20:26:58 BROKER 0: Nombre de Buffers de l'After-Image (-aibufs): 1. (4256)
20:26:58 BROKER 0: Nombre Maximum de Clients par Serveur (-Ma): 10. (4257)
20:26:58 BROKER 0: Nombre Maximun de Serveurs (-Mn): 13. (4258)
20:26:58 BROKER 0: Nombre Minimum de Clients par Serveur (-Mi): 7. (4259)
20:26:58 BROKER 0: Nombre Maximum d'utilisateurs (-n): 121. (4260)
20:26:58 BROKER 0: Nom du Host (-H): zeus. (4261)
20:26:58 BROKER 0: Nom de Name (-S): pp39. (4262)
20:26:58 BROKER 0: Type de Réseau (-N): tcp. (4263)
20:26:58 BROKER 0: Jeu de caractères (-cpinternal): iso8859-1. (4264)
20:26:58 BROKER 0: Stream (-cpstream): iso8859-1. (4265)
20:26:58 BROKER 0: Fichier Paramètre: d:\sytedb\reel\pp39_pc.pf. (4282)
20:26:58 BROKER 0: Port Minimum pour Auto Serveurs (-minport): 3000. (5648)
20:26:58 BROKER 0: Port Maximum pour Auto Serveurs (-maxport): 5000. (5649)

Sorry that's in french...tell me if you need me to translate

Thx
Mathieu

 

[Cringer]

Great stuff - that'll give the experts something to chew on. I'm sure the fact it's in French will make no bit of difference to any of them as the entries are undoubtedly in the same order as if they were in English

I'm guessing Syte is the reason you're still on an antiquated version of Progress? Or is there another reason? Because 8.3 is scarily out of date!!

 

[cj_brandt]

From the log it appears you are using the workgroup database. Is that correct ?

I'm going to guess the slow down is caused by the database starting a new server to handle the new connections. Is the database shutdown each night ?
The -Mi parameter allows 7 clients per server - when new client log in the database has to start new servers to handle the incoming connections. It appears from your settings you are using client server instead of shared memory connections. You could look at promon -> R&D -> 1 -> 3 before and after the slowdown to see if there are additional servers running.

Below are the parameters that control the launching of new servers.

20:26:58 BROKER 0: Nombre Maximum de Clients par Serveur (-Ma): 10. (4257)
20:26:58 BROKER 0: Nombre Maximun de Serveurs (-Mn): 13. (4258)
20:26:58 BROKER 0: Nombre Minimum de Clients par Serveur (-Mi): 7. (4259)

 

[Kenmath]

What do you mean by workgroup db ?

Yes all db are down during the nigth for the backup process...



Where is R&D in promon ?
Is there a way to change from client server to shared memory connections ?

Thx
Mathieu

 

[LarryD]

I could be wrong, but wasn't R&D a hidden option in promon in v8 ? Try typing 'R&D' in the 'enter your selection' just to check.

 

[Kenmath]

No you are a god !!
I'm in the hidden menu 1 -> 3 ok
So tomorrow morning I do a screen shot before and after the pb ok ?

Thx
Mathieu

 

[tamhas]

First thing that comes to mind is that your -Bs are tiny. What kind of hit ratio do you see in Promon?

 

[Rob Fitzpatrick]

A showcfg command in proenv will show you your installed licenses. The database will be Workgroup RDBMS or Enterprise RDBMS.

Where do all of your clients run? Are they on this box, or on remote machines? Or a mix of the two?

And since it's Windows, I have to ask: what else is running on this box?

What location does the S: drive map to? Is it a local drive, or remote?

How do your users authenticate? Against the OS, or do you have _User records in the database?

You are configured for a fairly high number of users. Do they each log in to the Windows server? How much RAM do you have in this box?

Is the Windows page file on the same disk as the database?

And while I'm at it... it looks like you aren't using after-imaging. You should be if this is a production system, unless data loss since your last good backup is not a concern.

 

[Rob Fitzpatrick]

First thing that comes to mind is that your -Bs are tiny. What kind of hit ratio do you see in Promon?

He has a 32-bit machine, and currently has 266 MB of buffer pool in his pp39 database (startup from the db log shows -B 34000 and 8K blocks). It appears this .pf is used to start clients, and it has some server parameters in it. So who knows what those DBs' buffer pool are actually sized to.

Based on my recollection of NT4-era boxes I used to run, there isn't likely to be a ton of RAM available here; I'd guess maybe 512 at most. Also, it isn't currently clear to me how many databases are involved, and which databases' broker or server processes are spiking the CPU. It would also be helpful to know if the offending _mprosrvs are brokers or 4GL servers.

 

[TomBascom]

I vaguely recall that Syteline may have some "daily cleanup" stuff that it tries to do when a user first logs in in the morning.

v8... ancient, obsolete, unsupported and pretty much begging to be put out of its misery The hardware it is running on is likely to be pretty bad too.

 

[TomBascom]

-n 121 = this must be an Enterprise license. Do you have more than 1 CPU? If so -- Magic trick #1, set -spin 10000

-Mn 13
-Ma 10
-Mi 7

After Magic trick #1 has been running for a day or two try:

-Mn 50
-Ma 3
-Mi 1

 

[Kenmath]

I vaguely recall that Syteline may have some "daily cleanup" stuff that it tries to do when a user first logs in in the morning.

v8... ancient, obsolete, unsupported and pretty much begging to be put out of its misery The hardware it is running on is likely to be pretty bad too.

Hi,
I'm the first connected all morning...and no connection pb...I'm doing some reporting during 10-15 minutes...server is busy because of that...but problem occurs after a couple of minutes ...

 

[Kenmath]

Dear all,
First I want to ty for all yours answers...
I made a word doc which contains all details of the "morning pb" with answers of all yours questions.
You can get the doc :
[url]http://dl.free.fr/rfwhrCqNH[/URL]

You have to enter the captcha words and then it will ask you for user/password :
No User
the pwd is : Kenm@th39

Thx for all
Mathieu

 

[TomBascom]

I cannot speak for anyone else but there is no chance at all that I am going to go through all of that to pursue these issues. I might go to the trouble of clicking on an attachment to a post. But even that is unlikely unless I had some reason to think it might be interesting.

I would much rather that you simply post your responses back to the forum.

 

[Cringer]

I can't even get the file when I tried.

If you want to send me the file (jdpjamesp AT gmail DOT com) I will host it somewhere more accessible for folks.

 

[Kenmath]

---

Sorry I understand...I thougth it was a good idea...
Below the document :







Server informations :
Please not that because of very old hardware, I virtualized this server.
Only bad point is that it has only one CPU now…instead of 2…but one good processor instead of 2 low
2 GB of RAM
C: 1.9GB – 0.8GB free
D: 23.4GB – 8.9GB free
Pagefile.sys is on D drive (1.3 Gb)
Showcfg : Entreprise DB (70 user limit)
Each client on the network runs this batch:
start /i S:\DLC83C\bin\prowin32.exe -basekey INI -ininame S:\reel\SymixClt-Local.ini -pf S:\Reel\pp39run.pf -p menu/symix.p -T %TEMP%
To run the SyteLine client
S: is a fix map drive of a share folder of the file server
Remote users are connecting to the TSE server (NT4) and run the batch as well.
On the syte server nothing else is running, just JetForm which is managing the printout from SyteLine.
Users authenticate according to _User records in the Syte db.

Hope this help...
Thx
Mathieu

 

[Kenmath]

I can't even get the file when I tried.

If you want to send me the file (jdpjamesp AT gmail DOT com) I will host it somewhere more accessible for folks.

mail sent
Thx Cringer