LH3 memory leak

[William Harrington_1]

Our LH3 print and file server PII 400hMhz with 1023MB of RAM (4 X256MB Kingston) loses all swapable memory after 5 days. We are now rebooting twice a week.
After five days the network access to the server is lost. When logging on to the console the machine is responsive but there are no services available.
Checking the event logs does not show any recorded event that would explain the lock up.
The network administrator indicated he saw the memory was totally used.
We are going to do a memory check with the memdoc boot disk this friday.
The load on this machine is fairly light, most of the printers have been moved off to newer servers.

Any ideas on curing this problem?

[kris rombauts]

Hi William,

i don't know the memdoc bootdisk test program , but is this is a diagnostic tool for memory problems, this will not help here as it sounds you have a memory leak issue at the OS level, this is not caused by bad memory chips but generally by bad s/w, driver etc.

If you run Windows, i would advise you to use performance monitor and monitor a couple of processes/objects and their 'private bytes' usage. You could even use task manager and enable some details for a start maybe.

Since the problem seems to reoccur after a few days already it must be a huge leak and by checking the perfmon data one should be able to filter out and narrow down (maybe it will take a few trials before you get their) to the module that keeps growing/consuming memory.

Also make an assessment of what changed last on your system and since when do you experience this issue, that could also make the link (i.e. s/w updates, newer drivers etc etc)


For H/W memory problems, look at the hardware log (the SEL) , if you have Netserver Toptools management agents installed then you can do this online. H/W memory module problems will never cause this kind of memory leak, they'll mostly simply cause a crash of the system but the problem then gets logged in the H/W log so one can replace the bad module.

HTH

Kris

[amhakassa]

Hi William,

This is what happen when you use a third party memory which is not actually made specifically to work with this server. The best bet will be to get rid of the kingston memory and replace them with HP memory.
The problem could go up to causing the systemboard to be unfunctional.

Regards
Amha Kassa

[HGN]

Hi

From what you have mentioned here looks like the server has been working for quite sometime with the kingston memory if that is the case then you need to try to test these modules on this server or someother server to make sure the meory is good,testing on this server can be done by removing/swapping memory modules .
If the memory was replaced recently then it is quite sure that this is related to the memory module.
Also you need to check if anything was done on the OS/software recently which may have also cuased this issue.

Rgds

HGN

[William Harrington_1]

Thanks for the comments:
The kingston memory has been in place for over a year and I wouldn't say every thing has been good since the upgrade of RAM but RAM has not been an obvious problem.
When I checked this morning by phone with HP parts they indicated that the largest RAM available was part number D5367-69001 at 128MB for $533.00 each or $395.00 each with orginal RAM trade in. I cannot justify $3000 for 512MB of memory. We're looking for the old RAM.
I found some RAM from a LPR server HP branded Part Number HP52R32952-86 PC100-333-620R. Would this be suitable for a LH3?

I also found reviews back in 1998 that indicated the LH3 was tested with four 256MB sticks so it seems odd that HP doesn't offer an upgrade at least 256MB. Any suggestions on what might be an acceptable and reasonably priced RAM?

[William Harrington_1]

I discussed with administrator to get more history.
We found the RAM that was in the LH3 was not HP, it was Apace which was upgraded by the HP people when we purchased the machine new.
In 2001 we upgraded fron NT to Win2000 server and at that time purchased the four Kingston 256MB DIMMS because Windows2000 would not run on just 512MB's.
This memory leak problem has just started during the last six months.

DOCMEMORY application is the one I'm using to test with and can be found at www.freedos.org

Thanks again for any comments

[William Harrington_1]

Finally had opportunity to do some serious work on this server. I found a hardware problem with the second drive cage. Even though the drives the full amount of space available the #15 slot is actually erroring out. I discovered this during a reconfiguring. This error would stop the adavance diagnostic utility from comleting it's attributes test.
I'm going to try a firmware update, if thatdoes not cure the problem I'll replace the drive.