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
