[Yum] [UG] yum, urlgrabber and no_proxy environment?

James Antill james-yum at and.org
Wed Dec 17 20:18:25 UTC 2008


Brian Long <brilong at cisco.com> writes:

> Let me summarize what I think you're suggesting:
>
> 1. Leave my environment alone (currently set via /etc/profile.d scripts
> for all users).
> 2. Install and enable James' plugin on machines that need yum no_proxy
> to work.
> 3. Copy the http_proxy line from /etc/profile.d/proxy.sh
> into /etc/yum.conf.
> 4. Add proxy=_none_ to specific repos which need no_proxy functionality.
>
> Is that correct?

 Not quite, if you are just setting "http_proxy=http://blah" in your
profile and you don't want yum to use that, then just install the
plugin.

 The main usecase for proxy=_none_ is if you have proxy="http://blah"
in yum.conf, but you want an internal only repo. to go directly
... then in the configuration for that repo. you set prox=_none_ and
then that repo. acts as if you hadn't set anything for proxy (and so
uses the env. variable, if present, or nothing).

> If so, I believe the only problem is that a mirrorlist would break this
> in my existing case.  I manage a private mirror of Fedora, Centos and
> EPEL.  Mirrorlist / MirrorManager sees my netblock and provides my
> internal mirror at the top of the list.  If that mirror is up and
> proxy=_none_, we're fine, but if my internal mirror is down and yum
> jumps to the second mirror, it will be unable to download anything since
> http_proxy IS required for the external mirror.  Is this correct?  If
> so, I think I can live with that over my current situation, but I just
> wanted to make sure I understand things before moving further along.

 You might be interested in reading:

 http://yum.baseurl.org/wiki/YumMultipleMachineCaching

> Is there any work in progress to get no_proxy working upstream in
> urllib2 / urlgrabber so I can turn off http_proxy per domain instead of
> per repo?

 You might want to create an upstream python bug against urllib2,
however it is very likely that yum/urlgrabber will move to using
pycurl in the next 6 months ... to fix this, and other issues.

-- 
James Antill -- james at and.org
C String APIs use too much memory? ustr: length, ref count, size and
read-only/fixed. Ave. 44% overhead over strdup(), for 0-20B strings
http://www.and.org/ustr/


More information about the Yum mailing list