Website monitoring - "socket timeout"

OK - so I'm not sure where this fits in the support system but firstly, GREAT to see the uptime monitor is working again! The bad news (for me), is that it has highlighted a major problem with my server with MANY outages due to "socket timeouts". This site is on a cloud VPS with 16GB of RAM and 6 'dedicated' cores. On a typical day I get around 40 - 45k visitors and 700k requests (see attached screenshots).

This is a new server, having moved off a dedicated server but I was assured by the new host that 8GB and 3 cores would be ample for this site. I doubled that for safety.

Would anyone like to venture an opinion as to the cause of these timeouts?

Note that the stats are from Cloudflare who claim their count is much more accurate than Google.

Thanks,

DW

    Luís

    Hi David ,

    Hope you're doing well today and thanks for your patience here!

    I am not sure if we can solve that issue increasing the "default_socket_timeout" variable:

    http://stackoverflow.com/questions/11682207/can-i-change-default-socket-timeout-from-my-php-code

    However, I already pinged the developers to get their valuable feedback. I will update this topic once I got a reply.

    Cheers, Luís

    David

    Hi Luis and thanks for your response. The challenge with increasing the timeout is that it only covers up the problem rather than fixing it. IWe have spent the past three or four days trying to work out what the issue is - without success.

    The site has gone from relatively fast to slower than 97% of all sites according to both Pingdom and PageSpeed.

    We've disabled plugins, rebooted the server etc., etc. This morning we'll disable the theme but it's the same theme we were using before the issue started and we can't find any comments from anyone else suggesting there is a problem with the theme.

    Looking forward to the dev's response.

    David Wilks

    James Morris

    Hello David,

    I'm sincerely sorry for the delay in replying to you on this issue.

    As you correctly stated, increasing the socket timeout only masks the problem. And the problem is optimization of the server to accommodate the demand being placed on it. There's also optimizing WordPress to accommodate the load.

    Plugin solutions such as Hummingbird and WP Super Cache and CDNs can only go so far. When you start hitting over 250,000 unique visitors per month, you will definitely have to dig under the hood of your server.

    Please keep in mind that just because a server has 16GB of RAM available and has SSD drives and has a massive network throughput, it doesn't mean it's been properly configured to leverage those resources. Ultimately, your software configuration has more to do with it than just the raw resources. Also, there's many little "tricks" that can be used to help squeeze every bit of performance out of a particular configuration.

    In order to give you some better direction on this, I need a bit more information. Can you please provide me with a general overview of your server configuration and what optimization steps you have performed on it?

    Best regards,

    James Morris

    David

    Hi James. I had given up on this post just as I gave up on the VPS. I moved back to a dedicated server a couple of weeks back and that has massively reduced our load times. I still get a couple of 1 minute outages every day (site only, not server) but two or three minutes out of 1,440 is acceptable. Just for reference, I'm running Nginx over Apache on a Xeon E3-1246 at 3.9GHz with 32gb RAM and SSD drives.

    The VPS was with HostNine who, hands down, offer the worst support I have ever experienced. It took days to get a response from a tech. The front line guys have very little real knowledge and even less authority.

    I went VPS because I had two relatively busy sites and a half-dozen small sites running off a single dedicated server. HostNine was recommended so - after many assurances from the front line people (all documented) - I bought THREE top level VPS servers and paid annually. Two months later I have now moved to InMotion and guess what... no refund nor even any apologies for the extraordinary lack of support.

    The lesson for anyone reading this is that it is absolutely essential you do your own checking before working off a recommendation. In this case, the referrer has only a couple of basic hosting accounts, is tech savvy so has not needed to rely on support. After experiencing all these issues I searched the forums for reviews on HostNine and was horrified by what I found. On the other hand, InMotion was consistently rated well.

    A very expensive lesson.

    Just saying.