[squid-users] Squid proxy not working when upgrade from 27 to 3.5

[Amos Jeffries]

On 25/10/18 10:19 PM, Angus J. wrote:
> 1. How is squid confdigured in windows (IE uses windows proxy settings)?
> NO
>  
> 2. whats's in squid access and cache logs? 
> -rw-r----- 1 squid squid      0 Oct 22 12:21 access.log
> -rw-r----- 1 squid squid      0 Oct 22 13:02 netdb.state
> -rw-r----- 1 squid squid   6498 Oct 24 11:29 store.log
> -rw-r----- 1 squid squid 141946 Oct 24 11:29 cache.log
> 
> 2018/10/24 11:27:34 kid1| Swap maxSize 102400 + 8192 KB, estimated 8507
> objects
> 2018/10/24 11:27:34 kid1| Target number of buckets: 425
> 2018/10/24 11:27:34 kid1| Using 8192 Store buckets
> 2018/10/24 11:27:34 kid1| Max Mem  size: 8192 KB
> 2018/10/24 11:27:34 kid1| Max Swap size: 102400 KB
> 2018/10/24 11:27:34 kid1| Rebuilding storage in /var/cache/squid (dirty log)
> 2018/10/24 11:27:34 kid1| Using Least Load store dir selection
> 2018/10/24 11:27:34 kid1| Set Current Directory to /var/cache/squid
> 2018/10/24 11:27:34 kid1| Finished loading MIME types and icons.
> 2018/10/24 11:27:34 kid1| HTCP Disabled.
> 2018/10/24 11:27:34 kid1| commBind: Cannot bind socket FD 24 to [::1]: (99)
> Cannot assign requested address
> 2018/10/24 11:27:34 kid1| commBind: Cannot bind socket FD 25 to [::1]: (99)
> Cannot assign requested address
> 2018/10/24 11:27:34 kid1| ERROR: Failed to create helper child read FD:
> UDP[::1]

Hmm, that is odd. I expect there is something wrong with the pinger
install and/or its security permissions.

But seems not to be having too much impact on the proxy. So looking into
it can be delayed to later.



> 2018/10/24 11:27:34 kid1| Configuring Parent 192.168.31.113/8001/0
> 2018/10/24 11:27:34 kid1| Configuring Parent 192.168.31.134/8005/0
> 2018/10/24 11:27:34 kid1| Configuring Parent 192.168.31.134/8004/0
> 2018/10/24 11:27:34 kid1| Configuring Parent 192.168.31.134/8000/0
> 2018/10/24 11:27:34 kid1| Squid plugin modules loaded: 0
> 2018/10/24 11:27:34 kid1| Adaptation support is off.
> 2018/10/24 11:27:34 kid1| Accepting HTTP Socket connections at
> local=[::]:3128 remote=[::] FD 18 flags=9
> 2018/10/24 11:27:34 kid1| Accepting reverse-proxy HTTP Socket connections at
> local=[::]:3128 remote=[::] FD 19 flags=9


Two http_port lines using port number 3128 ...

> 2018/10/24 11:27:34 kid1| Accepting reverse-proxy HTTPS Socket connections
> at local=[::]:80 remote=[::] FD 20 flags=9
> 2018/10/24 11:27:34 kid1| Accepting reverse-proxy HTTPS Socket connections
> at local=[::]:8000 remote=[::] FD 21 flags=9
> 2018/10/24 11:27:34 kid1| Accepting reverse-proxy HTTPS Socket connections
> at local=[::]:8004 remote=[::] FD 22 flags=9
> 2018/10/24 11:27:34 kid1| Accepting reverse-proxy HTTPS Socket connections
> at local=[::]:8005 remote=[::] FD 23 flags=9
...
> 2018/10/24 11:27:34 kid1| ERROR: listen( FD 19, [::] [ job2], 1024): (98)
> Address already in use

FD 18 and FD 19 both conflicting over who gets to listen on port 3128
and what type of traffic is arriving there.

This port is a registered port for forward-proxy use. Reverse-proxy
(accel mode) traffic has a *different syntax* - the URLs and types of
message that can be delivered are different. So cannot be sharing a port
with forward-proxy traffic.


Log says "ERROR" but is actually something FATAL. That is a bug we need
to fix in the logging and error display.


> 2018/10/24 11:29:31 kid1| Preparing for shutdown after 0 requests
> 2018/10/24 11:29:31 kid1| Waiting 30 seconds for active connections to
> finish
> 2018/10/24 11:29:31 kid1| Closing HTTP port [::]:3128
> 2018/10/24 11:29:31 kid1| Closing HTTP port [::]:3128
> 2018/10/24 11:29:31 kid1| Closing HTTPS port [::]:80
> 2018/10/24 11:29:31 kid1| Closing HTTPS port [::]:8000
> 2018/10/24 11:29:31 kid1| Closing HTTPS port [::]:8004
> 2018/10/24 11:29:31 kid1| Closing HTTPS port [::]:8005
> 2018/10/24 11:29:32 kid1| Set Current Directory to /var/cache/squid
> 2018/10/24 11:29:32 kid1| Starting Squid Cache version 3.5.21 for
> x86_64-suse-linux-gnu...


... and the auto-restart cycle continues.



Amos