Squid DNS FAIL when trying to connect to localhost - http

Question

I'm running a local http server and local squid instance. A local http client opens a socket connecting to the squid instance, which seems to work. I then try to tunnel to the local http server by issuing the following http request:
CONNECT localhost:80 HTTP/1.1\r\n
which yields the response headers
Content-Language en
Content-Length 3612
Content-Type text/html;charset=utf-8
Date Thu, 21 Jun 2018 17:28:10 GMT
Mime-Version 1.0
Server squid/3.5.27
Vary Accept-Language
X-Squid-Error ERR_DNS_FAIL 0
with status 503.
I also tried connecting to 127.0.0.1, which yields this response:
Content-Language en
Content-Length 3433
Content-Type text/html;charset=utf-8
Date Thu, 21 Jun 2018 17:35:16 GMT
Mime-Version 1.0
Server squid/3.5.27
Vary Accept-Language
X-Squid-Error ERR_CONNECT_FAIL 111
My squid.conf looks like this:
http_port 3128
coredump_dir /var/spool/squid
acl SSL_ports port 443
acl Safe_ports port 80 # http
acl Safe_ports port 21 # ftp
acl Safe_ports port 443 # https
acl Safe_ports port 1025-65535 # unregistered ports
acl CONNECT method CONNECT
acl any_host src all
acl all_dst dst all
http_access allow any_host
http_access allow all_dst
Is there a different way to tell squid to connect to localhost?

Answer 1

I found that what was failing what the localhost resolving to [::1] and not 127.0.0.1.
In order to bypass the /etc/hosts/ simply add the following to /etc/squid/hosts:
127.0.0.1 localhost
Then hosts_file /etc/squid/hosts in your squid.conf.
Of course the file can be put anywhere you would like.

Answer 2

Somehow squid tried to resolve localhost to 127.0.0.1, which ended up in a connection failure. Specifying [::1] instead of localhost, however, performs as expected.

Answer 3

In my case I was using the squid machine hostname (e.g. mysquid.proxy) and the problem was not related to the DNS resolutions because the squid machine could resolve itself correctly using its hostname.
The problem was rather caused by the configuration of an additional port in the same proxy. I was using squid as both forward proxy and reverse proxy with two different ports:
3128 - forward proxy
443 - reverse proxy
The client was connecting to the (forward) proxy mysquid.proxy:3128 and the request was something like:
CONNECT mysquid.proxy:443 HTTP/1.1
So the the reverse proxy port was used at the end.
However on that port it was configured a url_rewrite_program (a Perl script) to filter and change some path of specific url and such script was wrongly redirecting the request to a non-existant url which caused the error "503 Service Unavailable" in the client.