More information on configuring node connections.

Question

In the Netscape Calendar Server Admin Guide chapter on Managing a Node
Network, there is not sufficient information on configuring SNC connections.
The example provided shows "all:2" with no information on why 2 is a good number
for the example case. Other examples demonstrate increasing the number of connections
for specific nodes with no information on when this is appropriate.
<P>
We have configured our network with all:2, but we need documentation that details
how this number should be determined.
<p>
Two connections is the minimum in cases where there is one
node configured on each server. The unicwsd requires one connection for processing
of remote node requests and there should be at least one additional connection available
to be used by the clients.
<p>
Depending on how much traffic there is between nodes, two connections
might not be sufficient. The 'uninode -snc all' output should give you an
idea if there are not enough connections configured. An example of the output
follows:
<p>
uninode: connected to server1, node 10000
<br>Last reset: 29/6/99 13:00:51
PST8PDT
<br>----------------------------------------------------------------------------
<br><b>CONNECTION
EX CO AV US LOST RETRY QUEUE CANCEL CHECK GRANTED</b>
<br>----------------------------------------------------------------------------
<br>(10000)->server2(10001) 2
2 2 0
1
0
0
0
36
<br>(10000)->server3(10002) 2
2 2 1
0
0
0
5
126
<br>(10000)->server4(10003) 2
2 2 2
0
1 91
805
22
<br>(10000)->server5(10004) 2
2 2 0
0
0
0
0
181
<br>----------------------------------------------------------------------------
<p>In the example above, there are not enough connections available from
server1 to server4. <b>"US"</b> indicates that there were two connections
used at the time the uninode -snc all was run. <b>"QUEUE" </b>indicates that
there is one unieng process currently waiting for a free connection. <b>"CANCEL"
</b>means that there were x number of processes that did not get a connection, and
<b>"CHECK"</b> indicates the number of times a process had to retry before the request
for an snc connection was either granted or canceled. <b>LOST, QUEUE,
CANCEL, CHECK and GRANTED</b> are values that are reset when the unisncd
is shut down, or when 'uninode -reset' is run.
<p>
<b>Note:</b> In cases where there is more than
one node configured on one server, the node connections on that server going <b>to</b> the
other nodes are pooled. That means that any process on that server could use any snc
connection going to a node. In these cases the number of connections
configured could even be as low as all:1. Again, this depends on usage,
and might be subject to fine-tuning.

Answer 1

Hi,
you have to enable this additional measurements by executing
x_cons <dbname> time enable
Kind regards
Holger