Since there are no archives, I figured I would ask out here, since I can't
imagine that someone else hasn't seen this.

I have Sun E280R running Solaris 8, NV 7.1.3, Switch Analyzer 1.2, and
CiscoWorks 2000 with Campus Mgr, Resource Mgr Essentials, and Device Fault
Manager.

By default, the start order under Runlevel 3 is:
- /etc/rc3.d/S10dmgrtd (for CiscoWorks)
- /etc/rc3.d/S76snmpdx
- /etc/rc3.d/S99netnmrc (for NetView)

The interaction symptom is that something is causing snmpdx to delete the
connection to the mibiisa and mgragent subagents.  Once this occurs, it is
impossible to do SNMP queries to the box.    Worse, it is impossible to
restart the SNMP service, and subagents without rebooting.   

I suspect the culprit is Ciscoworks (or possibly Switch Analyzer)
- If I boot the machine normally, the SNMP stack is in the non-responsive
state described
- If I move the CW init script and bring up SNMP, and NetView, every is
working fine.   If I then start CW, then all 3 applications run fine for a
while.   However, after several hours it returns to this state.  
- Some combination of service restarts, also seems to cause this.

I have tried several things including:
- Using the "-f 0" flag on snmpxd
- Tracing snmpdx -d 4
- Adding a mgragent.acl, and adding a security stanza to mgragent.rsrc (as
per Sun support's instructions).

Despite debugging, the only meaningful messages have been posted to syslog.
A handful of SNMP authentication failures for queries originating from the
same box, session_open failures.   On some occasions, there will also be a
message about deleting snmpd or mgragent from the subagent table.

None of these fix the problem individually or in any combination that I have
tested.
I have PMRs open with all 3 vendors (IBM, Sun, and Cisco).    But, I'm
really starting to pull my hair out on this.

If anyone has seen a similar problem, please let me know.

Regards,

--
Chris Cowan
Sr. ESM Consultant
SRA International