Evgeniy Zaitsev wrote:
Hello.
We are using large squid cluster (~30 dedicated machines with squid) to
proxy-caching static content.
All separate squid's configured as sibling (each other, i.e. one level
hierarchy).
All requests to squid-machines go through one balancer. Balancer check
each squid (alive/not alive) via tcp-check (if port 3128/tcp accept
connections, then squid is alive).
But we want to use http checks for squid alive checking. Is it possible?
Depends on your balancer.
Squid does it automatically when "no-query" options is missing from
cache_peer lines.
A am search in this mailing list, but didn't found any way to do like this:
1. balancer create http request to squid (only _to squid_, not proxy
request to other http resource via squid) like "GET /ping" (or any other
HTTP GET)
2. squid anwer with currect http anwer code 200.
I found only one way to http check "squid alive" - balancer create http
request to another http server via squid, but this way is wrong for us.
If this external http server (for balancer check via squid) was died,
balancer reject all squids from cluster :(
Querying the cache manager interface should work.
cache_object://$SQUIDHOST/menu
should give you a 200 with the cachemgr menu options as data.
Any of the others should work as well, thats just the one with least
calculations done.
Amos
--
Please be using
Current Stable Squid 2.7.STABLE6 or 3.0.STABLE13
Current Beta Squid 3.1.0.5
