Re: [squid-users] Squid marks alive siblings as dead.

> On lör, 2007-10-13 at 02:00 +0100, Tony Dodd wrote:
>
>> > http://www.squid-cache.org/Versions/v2/2.6/cfgman/htcp_access.html
>> >
>> > It's likely deny by default.
>> >
>> > Chris
>>
>> Christ! In all my googling, and reading through squid books, not one
>> single page mentioned that. =/
>>
>> 2007/10/13 00:59:46| Detected DEAD Sibling: 10.0.12.8
>> 2007/10/13 01:00:08| Detected REVIVED Sibling: 10.0.12.8
>>
>> You are my savior Chris. Much appreciated.
>
> Please file a bug. HTCP replies are not supposed to be denied by
> htcp_access..
>

Henrik,
 I've looked at the code and I think this is caused as a side-effect of
"DEFAULT_IF_NONE: deny all" (@src/cf.data.pre:715) denying the initial
peer query (@src/htcp.cc:1236) when no other htcp_access are defined but
a peer is htcp_only.
I assume its there for security reasons, or to prevent unwanted HTCP sends
to non-htcp peers. But it is the only place htcp_access is ever checked.

I've already updated the .conf docs to clearly point out the htcp_access
needs to be explicitly configured for htcp peers.

Amos
Received on Mon Oct 15 2007 - 21:32:05 MDT