Re: [squid-users] 3.2.0.13: Cannot bind socket FD 21 to [::]:22381: (98) Address already in use

From: Amos Jeffries <squid3_at_treenet.co.nz>
Date: Mon, 24 Oct 2011 23:12:18 +1300

On 24/10/11 22:38, David Touzeau wrote:
> Le lundi 24 octobre 2011 à 10:40 +0200, Fred B a écrit :
>> De:
>> Fred B
>>
>>
>> Hi David,
>>
>> Please remove icp_port
>>
>>> Oct 24 09:35:32 prxysquid squid[3586]: assertion failed:
>>> icp_v2.cc:778:
>>> "Comm::IsConnOpen(icpOutgoingConn)" -> maybe ICP Bug with workers ?
>>
>>
>> ----- "David Touzeau" a écrit :
>>
>>> Dear
>>>
>>> I encounter many times this issue.
>>>
>>> "Cannot bind socket FD 21 to [::]:xxxxx: (98) Address already in
>> use"
>>>
>>> Perhaps it is a misunderstand of the SMP working process.
>>> I have 2 processors.
>>> Is each worker try to bind the same Port/IP ?
>>>
>>>
>>>
>>>
>>> Oct 24 09:35:31 prxysquid-squid squid[3586]: Starting Squid Cache
>>> version 3.2.0.13-20111023-r11381 for i686-pc-linux-gnu...
<snip>
>>> Oct 24 09:35:32 prxysquid squid[3586]: commBind: Cannot bind socket
>>> FD
>>> 21 to [::]:22381: (98) Address already in use

<snip>
>
> I had removed icp_port but the problem still occur :
>
> notice that i have
>
> http_port 22381
>
> perhaps it is an ipv6 issue ??
>

Should not be any more IPv6 port issues in 3.2 nowdays. Possibly one
left when sending v4mapped on a kernel with bind_v6only overrides, but
that does not show up in the bind() stage.

What does "netstat -antup | grep 22381" show?

If *any* IP on the box is already using it the wildcard will fail with
this message.

NP: I have seen this message happen on Linux if a previous server bound
the port then crashed. There seems to be a short timeout period where
nothing else is allowed to touch the port number.

Amos

-- 
Please be using
   Current Stable Squid 2.7.STABLE9 or 3.1.16
   Beta testers wanted for 3.2.0.13
Received on Mon Oct 24 2011 - 10:12:44 MDT

This archive was generated by hypermail 2.2.0 : Mon Oct 24 2011 - 12:00:03 MDT