[Search for users]
[Overall Top Noters]
[List of all Conferences]
[Download this site]
| Title: | DECmcc user notes file. Does not replace IPMT. |
| Notice: | Use IPMT for problems. Newsletter location in note 6187 |
| Moderator: | TAEC::BEROUD |
|
| Created: | Mon Aug 21 1989 |
| Last Modified: | Wed Jun 04 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 6497 |
| Total number of notes: | 27359 |
2396.0. "To many NML servers...." by MLNCSC::BARILARO () Mon Feb 24 1992 06:32
When I installed the MCC/BMS v1.1 in Provincia Autonoma di
Bolzano I was stunned by the number of NML processes created
on the customer's nodes from the Alarm rules, the Historian
and expecially from the EXPORTER.
For example:
The customer has a Vax 6420 with 6 async DDCMP lines, so when
I started the export directive for these entities (1 Executor,
7 lines/circuits Ethernet+DDCMP, 15 entities in total), the NML
processes from the MCC station pratically killed the network
connections to the 6420.
The "Maximum Logical Links Active" counter on the 6420 normally
between 5 and 10 reached the "Maximum Links" parameter of 32 and
need to be increase. This caused also problems on the alarm rules
because I used a rule that ask to a Decnet router (the same 6420)
if the remote node is reachable/unreachable (like the example
on ALARM manual), so when the alarm rule started they couldn't reach
the router because all his links were busy and I received the
Exception's mail.
After some checks seems that the EXPORTER open 4 (or more) links
for each of the 15 entities (one for the counters, one for the
characteristics, one for the status and one for the initial
attribute) for a total of 60 (or more, and I think more) links.
That topic is to ask if these observations are corrects, and if
there are a better way to implement the polling (for the Alarm
rules, Historian and Exporter).
BTW, for another problem (described in topic 2314) at the end in
Provincia di Bolzano I had to install ENOP for the alarming and
NMCC/Decnet Monitor to collect data, with these products I
hadn't this NML servers problem, they open a maximum of 3 links.
Thanks in advance for any help,
Luciano Barilaro
| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 2396.1 | See note 2213 for another way to poll DECnet nodes | CUJO::HILL | Dan Hill-Net.Mgt.-Customer Resident | Fri Feb 28 1992 00:59 | 17 |
| Every time you poll a DECnet node via the DNA4 AM, you will start
3 NML server processes. Two of these will eventually time out, leaving
only one left. This one will also time out when NETSERVER$TIMEOUT is
exhausted (default is 5 minutes).
If you are running two DECmcc polling processes from the same or
different Director nodes, you will get 3 NML processes created for each
DECmcc poll.
I created 24 NML processes on our cluster load host all at the same
time by using DECnet reachability alarm rules targeted at that node
in order to monitor 8 nodes. It almost crashed the cluster.
The problem is being worked and a solution will be made available
following testing.
-Dan
|