[squid-users] Accelerator http to https
creditu at eml.cc
creditu at eml.cc
Sat Nov 26 17:57:59 UTC 2016
Thanks for the explanation. We are on not on 3.2 (or greater) yet and
it doesn't appear concurrency is supported, so it looks like a single
threaded redirector for a little while longer.
On Sat, Nov 26, 2016, at 01:44 AM, Amos Jeffries wrote:
> On 26/11/2016 8:54 a.m., creditu at eml.cc wrote:
> > Using the first example in the link that was shared
> > (http://wiki.squid-cache.org/Features/Redirectors), I was able to get it
> > to work after seeing what was being sent to the redirector script. In
> > my case the URL was at $X and I had to remove all references to $X
> > in what was being sent back to squid. The below seems to work, but a few
> > questions since I would like this to be as robust as possible.
> The helper example uses concurrency to reduce the number of children
> needed. You enable that in squid.conf by adding "concurrency=N"
> parameter to the *_children line. With N being the count of requests you
> want it to handle at once.
> > If the presented URL is already https://..., my assumption is that just
> > sending back a new line is all that squid needs to see?
> Yes, an empty line for Squid-3.
> > Also, is all there any thing that I need to add besides the
> > url_rewrite_program and the number of children to the conf file?
> Just the concurrency level if you want to use that. see above
> > What
> > about turning off url_rewrite_host_header? The docs say this may be
> > wanted when running in accelerator mode. I did a few quick tests in a
> > test setup and don't see any difference.
> Nope, that is not relevant for proper redirectors.
> > Finally, is the best way to test how many children to launch (5, 10, 20
> > etc) just to monitor the cache.log to see if squid is running out and
> > increase it until the messages go away?
> > .....
> You could do it that way. Better way is to look at the cachemgr report
> about the redirectors. Each redirector should have a decreasing amount
> of usage, if you arrange the numbers for concurrency and children so
> that the Nth child has almost no lookups (at least less than its
> concurency N value) under your highest expected traffic load it is
> working fine.
> squid-users mailing list
> squid-users at lists.squid-cache.org
More information about the squid-users