Re: [squid-users] Failed to select source for - Unable to forward this request at this time

On 15/12/2011 11:46 p.m., karj wrote:
> Hi list,
> I am using Squid as HTTP accelerator, (Squid Cache: Version 2.7.STABLE9)
> and I've got the following problem-Error:
>
> ERROR
> The requested URL could not be retrieved
>
> While trying to retrieve the URL:
> http://swww.mgmt.example.com/themes/1/default/Media/Home/sylloges.arrow.righ
> t.active.gif
>
> The following error was encountered:
>
> * Unable to forward this request at this time.
>
> This request could not be forwarded to the origin server or to any parent
> caches. The most likely cause for this error is that:
>
> * The cache administrator does not allow this cache to make direct
> connections to origin servers, and
> * All configured parent caches are currently unreachable.

This error means exactly what its text says.

> At the cahe.log I've got the following message
>
> Failed to select source for
> 'http://swww.mgmt.example.com/themes/1/default/Media/Home/sylloges.arrow.rig
> ht.active.gif'
> 2011/12/15 12:14:48| always_direct = 0
> 2011/12/15 12:14:48| never_direct = 0
> 2011/12/15 12:14:48| timedout = 0
>
> At the access.log I've got
> 504 1648 TCP_MISS:NONE - - [15/Dec/2011:12:37:31 +0200] "GET
> http://swww.mgmt.example.com/themes/1/default/Media/Home/sylloges.arrow.righ
> t.active.gif HTTP/1.1" "-" "Mozilla/5.0 (Windows; U; Windows NT 6.1; en-GB;
> rv:1.9.2.24) Gecko/20111103 Firefox/3.6.24 GTB7.1"
>
>
> But I am sure that the URL is reachable in backend web server.

The error message happening means otherwise.

>
> The configuration options for the above URL are:
>
>
> refresh_pattern
> example\.com\/(.*)\.(jpg|gif|png|jpeg|css|js|axd|bmp|ico|swf) 1440 1000%
> 4320 reload-into-ims override-expire ignore-no-cache ignore-private
> override-lastmod ignore-reload
> refresh_pattern example\.com\/image\.limg 1440 1000% 4320 reload-into-ims
> override-expire ignore-no-cache ignore-private override-lastmod
> ignore-reload
> # First page 5 min
> refresh_pattern -i .example\.com\/$ 5 100% 6 reload-into-ims override-expire
> ignore-no-cache ignore-private override-lastmod ignore-reload
> # Articles page 6 min
> refresh_pattern -i .example\.com\/(.*)\/\?aid= 6 100% 6 reload-into-ims
> override-expire ignore-no-cache ignore-private override-lastmod
> ignore-reload
> # ArticleLIST Pagination page 60 min (articlelist/?pg=)
> refresh_pattern -i .example\.com\/(.*)\/articlelist\/\?pg= 60 100% 70
> reload-into-ims override-expire ignore-no-cache ignore-private
> override-lastmod ignore-reload
> refresh_pattern -i example\.com 1440 0% 4320 reload-into-ims
> override-expire ignore-no-cache ignore-private override-lastmod
> ignore-reload

I see a lot of nasty hacks going on in there. You are going to enjoy the
Surrogate-Control feature of Squid-3 when you upgrade.

Although why you would want to ignore and cache "private" objects in a
reverse proxy (which is shared by the entire Internet by definition) is
beyond me.

>
> #--- CACHE_PEER farm example start ----
> cache_peer xxx.xxx.xxx.xxx parent 80 0 no-query no-digest no-netdb-exchange
> originserver name=new_servers login=PASS
> acl www.mgmt.example.com_site dstdomain .exammple.com
> http_access allow www.mgmt.example.com_site
> cache_peer_access new_servers allow www.mgmt.example.com_site
> cache_peer_access new_servers deny all
> #--- CACHE_PEER farm example end ----
>
> icp_port 3130
> icp_hit_stale off
> log_icp_queries off
> icp_query_timeout 500 #MILI SECS
> visible_hostname KASANDRA.example.com
> cache_peer xxx.xxx.xxx.xxx sibling 80 3130 no-netdb-exchange no-delay
> no-digest proxy-only
> cache_peer xxx.xxx.xxx.xxx sibling 80 3130 no-netdb-exchange no-delay
> no-digest proxy-only
>
> I've the same configuration options on both siblings, all my servers have
> the same Problem.
>
>
> I've googled my problem and found that this happens when you run out of file
> descriptors which is not my case
> squidclient -p80 -h localhost mgr:info | grep 'file descri'
>
> Maximum number of file descriptors: 65534
> Available number of file descriptors: 62511
> Reserved number of file descriptors: 100

The filedescriptors is a only one of many paths which lead to a peer not
responding. It is also only relevant on the server to which connections
are failing (ie if the backend was a Squid which had run out of
descriptors). You are "dangerously close" to running out however.

"The Problem" is that Squid detected the peer down, but I think has no
way to detect it has become available again.

Squid requires 10 failed requests to detect a peer down, and just one to
detect it up again. These may be HTTP requests (normally none happen
while the peer is in 'DEAD' state), or ICP queries (disabled by
no-query), or NetDB checks (disabled by no-netdb-exchange), or cache
digest exchanges (disabled by no-digest), or ICMP pings (probably
disabled as well yes?).

To resolve that, in the case where the peer actually cant do any of the
services which are disabled, you can use the "default" option on the
origin server cache_peer line. That will make Squid attempt to use it
for HTTP request before declaring the error even if it is thought to be
dead.

Amos
Received on Thu Dec 15 2011 - 12:09:17 MST