Re: Peer twiddling

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Sat, 15 Sep 2001 18:32:10 +0200

Adrian Chadd wrote:

> WHat would the UNIX socket message format look like?

Text, command based.

> You won't be able to turn on a given peer:port via SNMP with only one
> SET- you'll need to walk a list of index->peer mappings, and then
> SET someoid.peeroid.port.enableoid={1|0}.

True, unless you index on something unique for the peer. Today peers are indexed by
their IP (had to change this in rproxy as I needed more than one peer per IP)

> That kind of thing requires some NMS magic - this'll come in the form
> of scripting.

The index -> named object lookup is fairly standard, and should be supported by all
NMS without any scripting. It is a very common operation when there are N of
something (interfaces, peers, VPN tunnels, ...)

> Personally, I don't think SNMP should be the default method of
> squid management. I think it should be HTTP - which allows it to
> inherit all the existing ACL types when it comes to management -
> and push out user-format-friendly and SNMP management via external
> agents.

I don't see SNMP as a good frontend to HTTP. The things you do with SNMP is most
often data collection, not management. With SNMP you only collect the entities you
want, but often more than one per collection interval.

I'd prefer if both cache_object and SNMP interfaces was as capable. But I do view
both interfaces as mostly a data collection tool for statistics and/or monitoring.
For now I'd prefer if active functions was done via another non-networked channel.

--
Henrik
Received on Sat Sep 15 2001 - 10:33:37 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:14:21 MST