[Setup] Re-post as: "Why Plone 4.2.3 page-loads slow if server not working hard?

[Héctor Velarde]

On 16/02/13 17:33, David Groos wrote:
>  1. First thing I need to do is set up more zeo clients.  I'm running
>     everything on an older quad-core xeon server w/6 GB RAM and only 1
>     client--never needed more before so never had to figure this out.
>     Looks like using only 1 client would explain under-utilization of
>     the hardware even though the web server is maxing out.
>      1. On this page:
>         http://developer.plone.org/reference_manuals/active/deployment/stack.html
>         it recommends one client per core--i.e. 4 clients for my server.
>      2. On this page:
>         http://developer.plone.org/performance/instancesthreads.html it
>         says the rule of thumb is 2 instances per core--i.e. 8 clients
>         for my server.  I saw this on another page.  Not sure what to go
>         for...

HV> those are only recommendations: you could start with 4 instances, 2 
threads per instance and 30,000 cache objects... you could need more ore 
less of everything, you have to figure out with the information from 
munin...

also, remember to install supervisor and make it restart your instances 
automatically when the memory consumption of them is above, let's say 
1GB: 
http://plone.293351.n2.nabble.com/best-practices-for-restarting-instances-td7459750.html

>  2. On this fairly recent page
>     (http://plonechix.blogspot.com/2011/08/10-minute-caching-with-apache.html)
>     Apache is recommended for the load balancing--will give it a try
>     since I already use Apache for rewrite to serve on port 80. Unless
>     someone strongly recommends something else?

HV> I would recommend nginx as it's smaller and faster than Apache.

take a look at this post from Nathan: 
http://nathanvangheem.com/news/nginx-with-built-in-load-balancing-and-caching

>  3. Munin (running on a virtualbox in which I'm running a plone site
>     from home) was using lots of cpu so I shut it down and haven't
>     persued it further.  At some point I'll probably get back to it...

HV> sorry to hear that; as I mentioned before, the only sane way to 
adjust your server parameters is by using munin; everything else is like 
shooting in the dark...

>  5. And in reference to Héctor's questions:
>         - how are they configured? how many threads? how many cache
>         objects? (Um, what is default ootb values?)

HV> IIRC 2 threads and 30,000 objects for Plone 4.x: 
https://pypi.python.org/pypi/plone.recipe.zope2instance

>  6. Can someone help me interpret the yslow info? But first, how should
>     I share that info--attach the index.html file? copy/paste onto a
>     pastebin page? Make some screen shots?

HV> just take a look at what is taking more time to load; this will give 
you a clue on where the bottleneck is...

regards

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 3758 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.plone.org/pipermail/plone-setup/attachments/20130225/248ac269/attachment.p7s>