Help!!! Cache Buffer Still Low after Server Memory Upgrade

[Arsynic]

Okay, I'm still learning Netware since I'm fresh out of college and landed a decent MIS job. I'm all smooth when it comes to our Windows Servers, but when it comes to the Novell stuff, we have a third party come in to handle the stuff (they set up the system before I came on board). One of the Master CNE's told me to watch the Cache buffer memory to indicate when we need more memory. Well, the cache buffer memory kept going down to 30% and 40% on a few of the servers. One of the servers is for a small health center of about 10 users. It originally had 384MB of RAM in it. I added 512MB more and it still drops over time?

Does this have something to do with garbage collection? If so, all of our servers are configured the same when it comes to that. One of our larger sites has the same configuration under Memory (Monitor --> Server Parameters --> Memory). The cache buffer memory stays at a rock solid 85% day in and day out since the server has been running. I have to do a server restart just to get the other ones back up to a respectable percentage, but they quickly go back down to the 40's and 30's. What do I need to do? I've been told by our third party MCNE that if the cache buffer memory gets too low, the server could abend. I thought adding more memory would do the trick, but it didn't. Does it have to do with processor speed as well? The speed of the server in question is a 860 MHz Pentium III. The server with 85% cache buffer memory is 1 GHz. I've been scouring the Novell Documentation, but to no avail. Help!

 

[marvhuffaker]

Adding more memory will not fix a leak. It might make it take longer to deplete completely, but it's not going to fix it.

Just from your description, it sounds as if something may be leaking but it's very difficult to tell without knowing more info about your server.

It would be helpful to know what NetwAre version, sp level, DS version, and what is running on the server. NDPS? Groupwise? Netshield? Btrieve? other stuff? If you could get that info, it will be easier to give you an idea of where to start.

You also MAY be able to see if there is ONE module that is consuming all the RAM. Go to MONITOR ==> System Resources ==>ALLOC Memory .... Sort by number in use. Find the one at the top of the list and see how much memory it is taking. Look at it after a reboot and then after the cache buffers goes down and see if it has changed significantly. you may find that this is a great place to start troubleshooting.

Marvin Huffaker MCNE, CNE
Marvin Huffaker Consulting

http://www.redjuju.com

 

[Lou0686]

Hi,

You will want to monitor the top ten modules listed on a daily basis, try to get some type of trend. Some modules will increase slightly in memory as time passes after a reboot. You may even notice that a module will creep into the top ten after a while. Anyway, the one that seems to grab the most over time and will most likely appear at the top of the list will most likely be your offending (leaking) module.

Lou

 

[Arsynic]

It would be helpful to know what NetwAre version, sp level, DS version, and what is running on the server. NDPS? Groupwise? Netshield? Btrieve? other stuff? If you could get that info, it will be easier to give you an idea of where to start.

Thanks a bunch guys, for your replies. First up, one of the main servers in question is running NW 5.1 DS version 8.79 and I'm not sure of the service pack level. The server runs Backup Exec Client and Job Manager, Symantec Antivirus Corporate Edition, Remote Console, and the timesync debug.

I checked the memory allocation and there's an NLM called
"MWMKDE" at the top that is eating up 191 MB of memory. Does that module sound familiar the KDE at the end sounds like some kind of development environment. If it is, I'm not sure if it needs to be running or not.

 

[Arsynic]

Oops...I read by the description that MWMKDE.NLM is actually a program that allocates memory. It seems as if it's allocating all the memory for itself. I wonder why that is?

 

[marvhuffaker]

Do you mean NWMKDE.NLM? That is Pervasive SQl (BTRIEVE), and it's generally used for basic server stuff. However, in your case, you are proabably running an app that uses a Btrieve database on your server. If that is the case, it's very possible to consume 191MB. I've seen some servers where it was in the 500MB range.

The key is to watch the NLMs over a few days and see if any of them start consuming an enormous amount of RAM.



Marvin Huffaker MCNE, CNE
Marvin Huffaker Consulting

http://www.redjuju.com

 

[Lou0686]

Hi,

NWMKDE is a Btrieve file that Veritas uses. The amount of memory that NWMKDE uses can be set in the BTI.CFG file. Usually when Veritas is installed, it will set the Cache Size to a large number and therefore NWMKDE seesm to be using lots of memory. You will notice that if you restart the server, the NWMKDE file will grab the same amount upon boot up. Althouh you can lower this amount, this is not your memory leak. The modules causing the leak will become more apparent as time passes after a rebbot. You will notice one growing much faster than the others.

Lou