Refreshing modified content

[Bedis 9]

On Fri, Dec 17, 2010 at 4:50 PM, Jonathan Leibiusky <ionathan at gmail.com> wrote:
> I think that conditional GETs could be really useful here. I just don't know
> how varnish works in this scenario. Maybe someone from the varnish
> development team can shed some light here :)
>
> On Fri, Dec 17, 2010 at 12:42 PM, James A. Robinson
> <jim.robinson at stanford.edu> wrote:
>>
>> On Thu, Dec 16, 2010 at 12:31, Paulo Paracatu <paulo at aliancaproject.com>
>> wrote:
>> > If I understood it, the purging method isn't automatic, right? I'd need
>> > to
>> > purge the content everytime it is modified.
>> > This is kinda stupid... I host more than 10k sites, modifying files
>> > everytime. If I set a high TTL, the backend will be happy and the
>> > webmaster
>> > will be angry. If I set a low TTL, the webmaster will be happy, but the
>> > backend will die. Plus, there is no point using a cache if the TTL is
>> > low.
>>
>> In a later post you ask about whether or not varnish could be
>> configured to send a conditional GET on every request
>>
>> Basically varnish would be looking up an item in its own cache, seeing
>> if it had a Last-Modified or ETag header from the backend, and sending
>> a conditional GET -- if it got an entity back it'd store that entity
>> as the new version, otherwise serve the old.  I'd be curious if
>> anyone's put together VCL logic that is capable of that.  It'd be good
>> to know how to do it.
>>
>> If it's possible to due, this technique might work well when you are
>> fronting a backend that is very fast at computing conditional GETs,
>> e.g., static files that can be examined to see if its inode, size, or
>> last modified time has changed.  I imagine most people using varnish
>> have slower backend servers, ones that build dynamic content and
>> aren't able to respond to conditional GETs any more efficiently than
>> they could respond to an unqualified GET.
>>
>> One of the places we use varnish at my dept is fronting a large (half
>> terabyte of about six million files) static file server.  Instead of
>> using some form of conditional GET, what we use are cache channels.
>>
>>  http://tools.ietf.org/id/draft-nottingham-http-cache-channels-01.txt
>>
>> When various programs are updating the backend server static files
>> they POST the filepath to our cache channel server.  We have another
>> program running that monitors the cache channel once a minute for
>> updates, and when it sees a new entry show up it turns around and
>> sends a PURGE request to varnish.
>>
>> Jim
>>
>> - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
>> James A. Robinson                       jim.robinson at stanford.edu
>> Stanford University HighWire Press      http://highwire.stanford.edu/
>> +1 650 7237294 (Work)                   +1 650 7259335 (Fax)
>>
>> _______________________________________________
>> varnish-misc mailing list
>> varnish-misc at varnish-cache.org
>> http://www.varnish-cache.org/lists/mailman/listinfo/varnish-misc
>
>
> _______________________________________________
> varnish-misc mailing list
> varnish-misc at varnish-cache.org
> http://www.varnish-cache.org/lists/mailman/listinfo/varnish-misc
>



Hi,

Maybe you can have a long TTL in your Cache-Control headers and force
a lower one into your VCL for Varnish itself.
my 2 cents :)

cheers