Re: [squid-users] Fail-Over Site Hosting

Philip de Souza wrote:
> Thank you, Amos. We're doing this as a one-off, but will repeat its use when
> similar maintenance requirements necessitate downtime in the future.
>
> For the stale-if-error statements, what does the actual statement look like
> in practice, please?
>
> There are no real case examples online yet from what I could find. Is it as
> simple as adding the line: "stale-if-error"?

Web server adds this header to all relevant requests:
   Cache-Control: stale-if-error

>
> Does one add a "> <milliseconds>" attribute at the end?

When the stale-if-error is no longer needed the web server needs to stop
sending it.

>
> Wouldn't "offline_mode on" work just the same?

offline mode in Squid is just a very aggresive caching mode. Not true a
offline.

>
> Also, what are the command lines needed in order to "triage" the pages in
> need of caching, please, as we definitely need this kind of control over the
> portions of our site that are to be kept available during the downtime?

If your cache is smaller than the website I'd run this for each URI in
the access.log going back at least a week:

    wget --mirror -p -l 1 -O /dev/null $URL

(may need some parameter adjustments t suite your needs).

Amos

>
> Many, many thanks,
> Philip
>
>
> -----Original Message-----
> From: Amos Jeffries [mailto:squid3_at_treenet.co.nz]
> Sent: Thursday, February 26, 2009 11:25 PM
> To: Philip de Souza
> Cc: squid-users_at_squid-cache.org
> Subject: Re: [squid-users] Fail-Over Site Hosting
>
> Philip de Souza wrote:
>> Hello,
>>
>> Can Squid-Cache be used in a Windows 2003 / IIS 6, web-hosting environment
>> to provide a temporary means by which the actual web server itself can be
>> brought down for maintenance, so that the site appears to remain live to
> any
>> clients wishing to access it?
>>
>
> Possibly.
>
>> We have two hosted machines with our web host provider, but were not able
> to
>> configure Network Load Balancing due to SSL cert.'s being needed on some
> of
>> the sites the second server is now hosting (thereby negating our ability
> to
>> use the 2 NICs on each machine we are limited to).
>
> Huh? SSL certs should not be needed for port 80 access...
>
>> The site we have running on the first server is a vital one, but the
> machine
>> needs maintenance and so we're searching for a solution that could work
> for
>> a period of 30-60 minutes, no more. Can Squid accommodate this, even for a
>> MS platform such as ours?
>>
>> Been searching all the documentation I could find and could not find the
>> answer I need on this. many thanks in advance!
>>
>>
>> ~Philip
>>
>
> Um, is this a once-off. Or a long term backup mechanism?
>
> Static content, or dynamic content that can be cached for the 30-60
> minutes without causing trouble is easy. Any purely dynamic content
> that cannot be cached at all may have trouble.
>
>
> To do this with Squid you would need version 2.7, with a cache large
> enough to hold most of the site and at minimum all of the in-use pages
> and files. Set squid up as a web accelerator for the machine thats going
> down. Then shift DNS over to it as primary server.
> http://wiki.squid-cache.org/ConfigExamples
>
> Once that is done the real prep can be started to enable caching of the
> site during the down time. Set up stale-if-error on all objects from the
> hosted sites so they can be served out of the cache even if too old
> during the downtime.
>
> Then use a spider (wget etc) to pull the important website(s) into
> Squid's cache fully just before bringing the web server down.
>
> With all clients going through Squid for some period leading up to the
> outage, there should be no major disruption.
>
> Problems you may face are:
> - not able to cache the whole website (triage the pages to those
> needed most)
> - dynamic pages not being changeable during the outage.
>
>
> Amos

-- 
Please be using
   Current Stable Squid 2.7.STABLE6 or 3.0.STABLE13
   Current Beta Squid 3.1.0.6