Hi,

I've just switched from OVH (small dedicated server) to a Linode 512 for performance issues. I'm really satisfied with Linode management panel and it's amazing how easy it is to create a new VPS server with backup within minutes. But I have performance issues detected with pingdom and Apache bench (ab).

On a client side, the website seems to run pretty well: http://www.escapade-carbet.com/

I used a stackscript (ID 360) to set up a Debian 6 with a LEMP Stack. The server is used to run a Wordpress website + W3 Total Cache + CDN (MaxCDN). The website files and backup have been transfered from a backup made with the Backupbuddy plugin. As I use frequently this workflow to migrate website, I'm pretty confident that performance issues don't come from the website files and database.

Pingdom is monitoring the site and throws repeatedly errors : "Connection refused HTTP CRITICAL - Unable to open TCP socket". Ping response times are not constitant, ranging from 100ms to 600ms. Pingdom errors occur more than 30 times per hour!

On another hand, Apache Bench resultats are not that good (running from another Linode server):

Requests per second:    35.78 [#/sec] (mean)
Time per request:       2794.491 [ms] (mean)

I tried to set up Varnish on a cloned Linode to improve performances, without any luck (messing up with VCL conf).

Any help would be appreciated. All these pingdom errors are driving me crazy!

Thanks!

A short update that could help people having the same issue:
'htop' command was showing a lot of swapping.

I've just noticed that pm.max_requests was still commented in /etc/php5/fpm/pool.d/www.conf

After a few PHP-FPM tweaks:

pm.max_children = 8
pm.start_servers = 2
pm.min_spare_servers = 2
pm.max_spare_servers = 4
pm.max_requests = 100

And in /etc/php5/fpm/php.ini

memory_limit = 64M

Now ab is running smoothly (more than 70 req/s), without swapping on server side. And pingdom outages stopped.

So, for people using stackscript #360, don't forget to uncomment "pm.max_requests" after server deployment.