Has anyone else experienced memory problems running ND6 on a Windows 2000 Server (Service Pack 3)? The following executables seem to grab on to memory and hang on for dear life:nserver.exe
nupdate.exe
nadminp.exe
nrouter.exe
In previous releases, various executables would spike in memory usage when heavily utilized but would later release memory. In ND6, the memory just grows and grows until the server is unreachable. We’ve escalated an incident with IBM to the highest level possible, but they’ve been unable to give us anything to go on. Judging from the comments others are posting, it seems pretty much everybody is happy with ND6’s reliability (compared to the initial response when R5 was released); we seem to be the only ones having this problem. Additionally, this is happening on every single one of our servers, even a mail server supporting only 80 users. Any ideas?
I have upgraded our mailserver to 6.03 with W2kSP4 and the memory leak problem remains. nserver & Co uses all memory they can find, but releases it about once a day. That won’t do because server is very slow all the time.
We have other 6.02x servers with the same problem. Availability Index is below 20 and CPU usage average is less than 15. So IBM do something because you have couple of thousand angry customers here…
I have removed GroupShield for Domino and that seems to have helped the problem a lot. We’ve had 3 weeks of uptime, the memory usage seemed to stabilize for a few days at around 830 MB, but we have seen an increase in memory usage of about 50 MB in the last week. I’m still waiting for the server to crash again so I can NSD and get further analysis from Lotus Support. Do you have GroupShield or any other 3rd party apps running on your server as well?
I have also Groupshield for Domino 5.2.1 running, but stopping it doesn’t help the speed problem. When there are 300 people using standard mail5 and 250 inotes5(web) users, memory usage is about 1GB total but task manager tells that nserver & Co uses over 4GB. ;-)That can’t be right, because server still has 500mb free memory and it isn’t swapping.
I have nothing but mailboxes on server and it’s clean install. ND5.0.11 was a lot better. Maybe I should downgrade…
Stopping GroupShield didn’t help us either. We had to actually remove it to stop the crashing. If you discover that removing GroupShield helps you as well, check out www.appriver.com . We outsourced or virus and spam filtering to these guys when we removed GroupShield, and for $1500 a year, it’s more effective than GroupShield and Domino 6’s blackhole list combined… Especially when you consider the whole Osirusoft fiasco a couple months back.
And as for Taskman… We’ve discovered elsewhere in this thread that it’s a big fat liar. Right now our total memory usage is about 880 MB, but if you add up the individual processes the total is more than 3 GB, which is more than our physical RAM and swap space total.
Try this little experiment: Minimize your Domino Server console and restore again. Voila! Your memory usage for each individual Notes task has dropped to just a few MB each, but the total is still the same. Need any more proof that Taskman is a liar?
My scenario is this , Our domino server, serves 40-50 users.
Specs are,
Domino 6.0.2CF2
AMD 1800 XP
768 PC2100 DDR
W2k Pro SP4
What has happened since these notes.ini entries.
Re-started domino server this morning, nserver.exe started to increment etc.
By lunch time, domino was using roughly 500MB (nserver.exe 215MB) of total system memory. Came back from lunch an hour later nserver.exe was down to 30MB, task manager said I had 320MB physical RAM back.
No crashes while I was at lunch, no nsd log etc.
No orderley shut downs etc, checked server logs, statistics etc.
So it must of got to approx 550MB and decided to free up resources ?
We’re running into the same problem. NSERVER.EXE and NROUTER.EXE hog up the CPU constantly. A server reboot fixes the problem temporarily. Quick Editmode on console is turned off. Any other suggestions?
“Try this little experiment: Minimize your Domino Server console and restore again. Voila! Your memory usage for each individual Notes task has dropped to just a few MB each, but the total is still the same. Need any more proof that Taskman is a liar?”
This remark of Rick exactly match my issue and minimizing and maximizing changes a lot…
Could you people also check this?? Allthough it seems TaskMan is a liar, after a few days when the domino processes are about 800MB(!!) the server starts to get slow anyway… After using the above solution the server starts to respond normally…
Try running your server without the Domino Console. I have just started testing this myself, but it seems to be working so far.
I’m running Windows 2000, SP4 and Domino 6.5.1. I have Domino running as a Windows service. I noticed the same behavior on my server that is documented in this thread.
To run Domino without the console, open regedt32. Browse to \HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Lotus Domino Server (LotusDominoData). Change the ImagePath entry to include “-jc -c” after nservice.exe. So my ImagePath entry ended up looking like:
I’m still having memory issues since making the change. It’s definitely better, but not resolved.
Since that didn’t solve it, I began restarting Domino every morning at 5am. I created a batch file that did a simple net stop, then a net start of the Domino service. That worked for a little while. Then my tempermental server started having issues where the net stop portion of the batch file wouldn’t stop the service. When it tried doing the net start, it wasn’t able to start Domino because the server thought it was already started. Domino would end up in some half started/stopped state. The only thing that seemed to resolve it was a server restart.
Now, we’ve scheduled a server reboot that takes place 4 times a week (M,W,F,Su) about 30 minutes before our nightly backup. It seems to be working for now, but I hate the fact that I am forced to reboot this often. I should be able to leave it up and running for months at a time!
Has this change resolved your problem? Also, how do these switches(-jc -c) differ from just using the checkbox in services “Allow service to interact with desktop”.thanks,
I can say for a fact, that with 6.0.1CF1, adminp will gobble up about 500K-1mg per database when operating on a “Change name in ACL” request and never give it back.
Usually not a problem, until you have a server with 5,000-6,000 databases.