fidzu : plone

09 Nov 2011 9:30pm GMT

This just in from the production department: use Varnish. (And please forgive the heavily meme-laden approach to describing these techniques .)

Cache ALL the hosts

Our ability to use Varnish in production is no secret by now, or at least it shouldn't be. What is often less clear is exactly how to use it. One way I like[1], is to run Varnish on your public IP port 80 and make Apache listen on your private IP port 80. Then proxy from Varnish to Apache and enjoy easy caching goodness on all your virtual hosts in Apache.

Configuration

This should require less than five minutes of down time to implement. First, configure the appropriate settings. (Well, first install Apache and Varnish if you haven't already: `aptitude install varnish apache2` on Ubuntu Linux[0].)

Varnish

To modify the listen IP address and port, we typically edit a file like /etc/default/varnish (in Ubuntu). However you do it, configure the equivalent of the following on your system:

DAEMON_OPTS="-a 174.143.252.11:80 \
             -T localhost:6082 \
             -f /etc/varnish/default.vcl \
             -s malloc,256m"

This environment variable is then passed to varnishd on the command line. Next, pass traffic to Apache like so (in /etc/varnish/default.vcl on Ubuntu):

backend default {
 .host = "127.0.0.1";
 .port = "80";
 }

Now on to Apache.

Please note that the syntax above is for Varnish 3.x and the syntax has (annoyingly) changed from 2.x to 3.x.

Apache

The Apache part is a bit simpler. You just need to change the listen port (on Ubuntu this is done in /etc/apache2/ports.conf), typically from something like:

Listen *:80

to:

Listen 127.0.0.1:80

Restart ALL the services

Now restart both services. If all goes well you shouldn't notice any difference, except better performance, and when you make a website change and need to clear the cache[2]. For this, I rely on telnetting to the varnish port and issuing the `ban.url` command (formerly `url.purge` in 2.x):

$ telnet localhost 6082
Trying 127.0.0.1...
Connected to localhost.
Escape character is '^]'.
200 205     
-----------------------------
Varnish Cache CLI 1.0
-----------------------------
Linux,2.6.35.4-rscloud,x86_64,-smalloc,-smalloc,-hcritbit

Type 'help' for command list.
Type 'quit' to close CLI session.

ban.url /
200 0

Cache ALL the disks

This site has Varnish and Apache configured as described in this article. It also has disk caching in Apache enabled, thanks to Elizabeth Leddy's article:

• http://plonechix.blogspot.com/2011/08/10-minute-caching-with-apache.html

As a result, it's PEPPY AS THE DICKENS™ on a 512MB "slice" (Cloud server) from Rackspace Cloud. And now you know yet another "Plone secret". Now go make your Plone sites faster, and let me know how it goes in the comments section below.

Notes

[0] Using the latest distribution, "oneric".

[1] I first saw this technique at NASA when NASA Science was powered by Plone; I found it odd at the time but years later it makes perfect sense.

[2] Ideally you'd configure this in p.a.caching, but I've not been able to stomach this yet.

09 Nov 2011 5:50pm GMT

Just add the method below to the RedirectionTool and call it from the browser as:

http://localhost:8080/site/portal_redirection/getAllRedirects

Assuming that the site is running at loaclhost:8080 that is :-S

That will show a list of redirects that can be imported into plone 4.x


security.declareProtected(View, 'getAllRedirects')
def getAllRedirects(self):
"get'm'all"
result = []
reference_tool = getToolByName(self, 'reference_catalog')
for k,uuid in self._redirectionmap.items():
obj = reference_tool.lookupObject(uuid)
if obj is None:
print 'could not find redirect from: %s to %s' % (k, uuid)
else:
path = '/'.join(('',)+obj.getPhysicalPath()[2:])
result.append( '%s,%s' % (k,path) )
return '\n'.join(result)


08 Nov 2011 2:58pm GMT