On Tue, 16 Aug 2011 12:09:27 +1200, Rhys Evans wrote:
On 25 May 2011 19:17, Amos Jeffries <squid3@xxxxxxxxxxxxx> wrote:
On 25/05/11 14:10, Rhys Evans wrote:
Seeing as I don't have any neighbour caches, I set htcp_port 0 and
things seem to be OK. It still seems odd that this issue came up,
I'm
wondering if maybe some Windows update altered some behaviour in
regard to allowing the binding required?
Looks similar to the Linux problem when Squid is not started as root
(no permission to open its listening sockets).
Since it was working before, there may be some registry change
around permissions involved. Possibly from the update or corruption
from eth crash you mentioned. Figuring out what and how is the tricky
part.
Amos
We've just had a few more power cuts and also more updates. A new
error came up which stopped squid from starting.
The error is:
commBind: Cannot bind socket FD 15 to *:3130: (10013) WSAEACCES,
Permission denied.
FATAL: Cannot open ICP Port"
I set ICP to 0 and it now seems to be running OK but I'm not sure if
that's desirable and it just seems strange that a config will work
for
a while and then fail. I couldn't find out much else on the topic but
maybe I'm not searching for the right terms.
I suspect its general security access to bind() on high numbered ports.
For both ICP and HTCP.
The server or group security policies will affect these.
Side effects of disabling both will be an impact in external bandwidth
and duplicated content in the sibling caches. If you used it for parent
selection then this could also lead to mistakes in their up/down
detection, resulting in more client visible error pages. Hopefully not a
major problem, but annoying.
Amos
