Perhaps these are two separate issues; perhaps they are related.

Quick scenario:

Proliant DL380 G4, 4GB RAM, OES NetWare SP2 (NW65SP5), BM 3.8 SP4 with
IR5 applied
Several third-party S2S VPN tunnels up
Several C2S Skip VPN tunnels up
Proxy Auth configured
All http and https (via port 443) traffic proxied from remote locations
as well as LAN

After about two weeks, I got a call that certain sites were unreachable
via proxy. It turned out that the C2S connections were getting dropped
and had to reconnect (not sure of the S2S connections). While
troubleshooting the problem over the phone with the client, I found that
while LAN users (including Admin) could authenticate to the server,
drives were not getting mapped. Users already logged in were able to
continue functioning.

My own network is configured as a S2S VPN from NSM (my side) to the BM
server. My connection was idle until a few minutes ago (24 hours after
the first trouble report), when I attempted to examine the DHCP
assignments with the DNSDHCP Java console. Partway through my browsing
of address assignments, I received an error. Looking further, I saw that
while my VPN connection was still up, I had no IP traffic able to make
it through the tunnel.

Using the remote insight board, I opened a remote console to the server
to find that nothing had changed; it was just where I'd left it a few
moments ago (no abend, nothing in the console logger). I stopped my VPN
connection from my end, and re-started it (using the NSM web insterface)
and my traffic started flowing again. A couple minutes later, after
similar activity in DNSDHCP, the same thing happened. This time, I did a
quick reinitialize system, and again restarted the VPN from my end of
the tunnel. Things are now back up and seem to be staying that way.

Next, I attempted to view the contents of my SYS:NWPATCH directory just
prior to this post, to verify that I did indeed have an installation
repository for IR5. TOOLBOX.NLM came back with an 8901 error when trying
to authenticate to NDS. DSREPAIR comes up clean when running in
unattended mode and when rebuilding the local database (I did not
rbeuild the operational schema, however, I don't think that's an issue,
here). I can't unload LSAPI.NLM to attempt reloading it without
disrupting the VPN services (I can do that tonight, however).

Might these issues be related to some buffer overflow? System Resources,
from the console monitor, show:

Allocated memory pool, in bytes: 1,237,524,704 33%
Cache buffer memory, in bytes: 1,062,559,744 28%
Cache movable memory, in bytes: 0 0%
Cache non-movable memory, in bytes: 4,833,280 0%
Code and data memory, in bytes: 98,579,841 3%
Miscellaneous memory, in bytes: 1,353,968,255 36%
Total server work memory, in bytes: 3,757,465,824 100%

Proxy Memory Usage Statistics:

Request Processing: 458,752 bytes
Connections: 114,688 bytes
DNS Cache: 733,184 bytes
Object Cache: 634,880 bytes
ICP Client and Server: 0 bytes
Other: 2,605,056 bytes

Total Allocated Memory: 4,546,560 bytes

The only real memory tweaking I have done on the server was to increase
the Java heap so that my web apps would remain functional. To do this, I
added the following in AUTOEXEC to reserve a gig:

SERVER -U=10000000000

and then, in tomcat4.ncf:

setenv CATALINA_OPTS=-Xmx900m {remaining params}

to remove the minimum heap size and increase the max to 900MB.

At some point I'll probably lower the initial reserved amount to 900MB
and get back the extra 100MB which is now unused.

Anyway, any thoughts as to what to check next?

Lewis G Rosenthal, CNA, CLP, CLE
Rosenthal & Rosenthal, LLC
Accountants / Network Consultants
New York / Northern Virginia
eComStation Consultants
Novell Users Int'l
Need a managed Wi-Fi hotspot?