Re: [squid-users] Best Configuration for sibling peer

From: Kris <christian_at_wanxp.com>
Date: Thu, 17 Dec 2009 13:15:12 +0700

Chris Robertson wrote:
> Kris wrote:
>> Chris Robertson wrote:
>>> Kris wrote:
>>>> Dear All,
>>>>
>>>> i have 4 proxy server with about 1000 request per second average ,
>>>> i have extra free nic in every server and i connected that 4 proxy
>>>> to 1 switch and give them 1 network ip. i set 4 proxy as SIBLING
>>>> each other. after few days i got problem like.
>>>>
>>>> 1. weird error log (sometimes)
>>>>
>>>> 2009/12/16 09:48:25| storeClientReadHeader: no URL!
>>>> 2009/12/16 09:48:25| storeClientReadHeader: no URL!
>>>> 2009/12/16 09:48:25| storeClientReadHeader: no URL!
>>>> 2009/12/16 09:48:28| storeClientReadHeader: no URL!
>>>> 2009/12/16 09:48:29| storeClientReadHeader: no URL!
>>>>
>>>> 2. TCP Connection Failed
>>>>
>>>> 2009/12/16 09:45:15| TCP connection to 10.10.10.10
>>>> (10.10.10.10:3128) failed
>>>> 2009/12/16 09:45:22| TCP connection to 10.10.10.11
>>>> (10.10.10.11:3128) failed
>>>> 2009/12/16 09:45:22| TCP connection to 10.10.10.11
>>>> (10.10.10.11:3128) failed
>>>> 2009/12/16 09:45:28| TCP connection to 10.10.10.13
>>>> (10.10.10.13:3128) failed
>>>> 2009/12/16 09:45:36| TCP connection to 10.10.10.12
>>>> (10.10.10.12:3128) failed
>>>>
>>>> 3. sometimes it become hard to browse , but it going normal after i
>>>> disabled all SIBLING, if i enabled it will slow again.
>>>>
>>>> my file descriptor already 65535
>>>>
>>>> my peer config
>>>> # cache_peer 10.10.10.10 sibling 3128 3130 no-netdb-exchange
>>>> no-digest no-delay round-robin proxy-only
>>>> cache_peer 10.10.10.11 sibling 3128 3130 no-netdb-exchange
>>>> no-digest no-delay round-robin proxy-only
>>>> cache_peer 10.10.10.12 sibling 3128 3130 no-netdb-exchange
>>>> no-digest no-delay round-robin proxy-only
>>>> cache_peer 10.10.10.13 sibling 3128 3130 no-netdb-exchange
>>>> no-digest no-delay round-robin proxy-only
>>>>
>>>> any suggestion what best configuration for sibling peer ?
>>>
>>> I don't think that round-robin makes much sense with sibling
>>> selection. Either the a sibling has the object (and we grab the
>>> object from it) or it doesn't (and we grab the object directly).
>>>
>>> For myself, I use a line like...
>>>
>>> cache_peer proxypool-2.my.domain sibling 8080 3130 no-digest
>>>
>>> Chris
>>>
>>>
>> did you ever get message like " TCP Connection Failed" ? just curious
>> why i always got that message , peer connection use same network in
>> same switch.
>
> My peers use the same interface to talk to clients and each other. I
> don't see "TCP Connection Failed" messages. Hmmm... One thing you
> might try...
>
> acl myPeers src 10.10.10.0/24
> tcp_outgoing_address 10.10.10.10 myPeers
>
> ...to make sure it responds to the peers using the right interface.
>
> Chris
>
>
>
i have 1000 request hit per sec in every proxy , is that can be a cause ?
Received on Thu Dec 17 2009 - 06:15:21 MST

This archive was generated by hypermail 2.2.0 : Thu Dec 17 2009 - 12:00:02 MST