RE: [squid-users] LVS & cachemgr/snmp?

From: apiesk <apiesk@dont-contact.us>
Date: Tue, 4 Sep 2001 12:54:16 +0200

> From: mkiernan@mail.onet.pl [mailto:mkiernan@mail.onet.pl]On Behalf Of
> Mike Kiernan
> Sent: Tuesday, September 04, 2001 12:11 PM
> To: squid user group
> Subject: Re: [squid-users] LVS & cachemgr/snmp?
>
> I suspect I'm talking to myself again here
> [this list gets awful lonely sometimes :) ]

maybe many people are on vacation ?

>
> There doesn't seem to be an easy way around this.
>
> Looking at the snmp code, it returns a
> barely related set of stats which don't
> contain the cachemgr subset I'm after.
>

yeah i wish the snmp-mib would contain the whole cachemgr set.

> The only way to suck in the cachemgr
> stats is through the main squid port.
>

so why don't you get the stats directly from the nodes.

point cachemgr.cgi to node1 for instance.
this assumes you have access to the single nodes and
the squid is listening on the host ip.

if you have no ip-access to the single nodes (only to the cluster-ip),
i would set up redirection using iptables.

lvs-ip:8001 -> node1
lvs-ip:8002 -> node2
etc.

but maybe i totally missed the point.

ciao -ap

> slightly frustrating, but I'll live.
>
> - Mike
>
> > One of our squid farms looks like this:
> >
> > cluster1.pl[public]:{[private] node1 node2 node3 }
> > |
> > | cluster2.pl[public]:{[private] node1 node2 node3 }
> > | |
> > |{control workstation:cachemgr.cgi}
> >
> > Each node has 2 squids running on it, one
> > cacheing, one redirecting/delaying.
> >
> > I want to setup the control workstation so
> > it sucks in all the 12 cachemgr stats to one
> > central location (we have no webservers within
> > the squid farm to server cachemgr.cgi).
> >
> > Problem is the cluster is LVS (linux virtual server)
> >
> > It's thus not possible to reliably obtain correct stats
> > for each node because of the round-robin. I can
> > probably work around by proxying the cachmegr request
> > to another port eg: node1:81 where node1 is an internal
> > static ip and forwarding it locally on the machine to
> > cluster1.pl:80, but this seems a bit ott.
> >
> > Any way squid can dump this data to another port?
> >
> > I haven't looked at any of the snmp stuff yet - does this
> > return the same stats - is it a viable alternative
> > to grab the data?
>

System Administration
VIRBUS AG
Fon +49(0)341-979-7424
Fax +49(0)341-979-7409
andreas.piesk@virbus.de
www.virbus.de
Received on Tue Sep 04 2001 - 04:54:20 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:02:01 MST