I've hit a (server)brick wall

MrMedia

MrMedia

Joined
Mar 27, 2015
Messages
827
Likes
1,476
Degree
3
This is baffling to me.

I have a site hosted on a VPS on Liquidweb.

My site gets about 25k visitors a day.

However when traffic on site peaks around 200 people on site per minute the whole site either slows down terribly or crashes and becomes non responsive.

This seems bizarre to me since its such a small amount of traffic.

Below is the response from Liquid web support.

As a server noobcake I have zero clue what any of this means of if they are just bullshitting me and I should move host.

Anyone technical have a clue what they are on about and what I should do?



That means that the domain xxxxx.com is getting enough traffic that its causing load and bringing down sites. I found the cause of this to be in the .user.ini in the public_html directory for that user. There is a PHP memory limit there of 196M. With that PHP memory limit, there is a maximum possible usage of 14.7G of RAM and there is only 3.3G available on the server.

One way to alleviate that issue would be to raise the RAM on the server to at least 15G. Alternatively we could lower the PHP memory limit for that .user.ini.

Please let me know how you would like to proceed.

my recommendations mean we will want to either lower the PHP memory limit for that site OR increase the RAM on the server to accommodate that site OR lower the Apache setting for simultaneous connections to accommodate the PHP memory limit for that domain.

The server is not broken by any means. You would want a developer to fix the code. Also they can make recommendations or clarify why sites are using so much resources.
 
As a long time LiquidWeb customer, this is pretty common. What kind of site is it? (yes platform, third party code on top, users, etc...)

It's pretty much standard procedure for them to say have your developer mod code, blah blah... implement a request limit at some point, so consider yourself lucky you have the opportunity to fix this... instead of waking up one day wondering why your traffic is cut in half, only to realize your box has been throttled to only allow so many connections per second. (this is usually a problem for any site running a db, without aggressive caching)

Doesn't make them a bad hosting company, and I still have a server there, but it's like going to Best Buy looking for a custom gaming setup... you're at a big box store, wanting services/products outside of the box at times. The trouble is they cater to a WIDE variety of customers, so they're configurations (and support staff) are calibrated for that. Typically they will recommend that you need to upgrade, until suddenly you find yourself on a $600/mo dedi, and are having the same fucking problem as day one. The tier 1 techs are usually the culprit here, and you have to know the right questions to ask before you get anywhere meaningful.

Step 1) site details (type, code, db?, ad code, third party scripts, waterfall screenshot, and how many pages indexed: you want to make sure you don't have a crawler issue if this is a really large site, and it's not actually user sessions bringing your instance to its knees)
Step 2) cache or no cache
Step 3) optimize
Step 4) upgrade, migrate, or not
 
200 people on a site per minute multipled by 196MB = 39GB of RAM needed to sustain those users, (you are hitting 14 GB since it doesn't take a whole minute to process data). Lower the PHP memory by their recommendation. As well increase the RAM on the server, even if you were to half the RAM used you still would need 18.5GB of RAM for your server setup. You have only 3.3GB.

When you don't have enough RAM on your computer and your computer needs to still process all that data the computer starts using your HardDrive to store memory. Your HardDrive is more than 2000x slower than RAM. That is why the website slows down.

Basically you have a website that is an SUV car, that can fit 8 people, but 200 people are piled into it... you can't go the same speed, and there is a lot less room with 200 people in your car designed for at max 8 people. You clearly need a bigger car and more efficient - you need a train for 200 people per minute.

If you are getting 25K a day in visitors a plan with 3.3GB of RAM - that is way too small. Upgrade to 15-20GB if possible, reduce the PHP memory (after asking the web developer why it was set to 196MB that PHP can used might be a reason) and you'll be fine.

They aren't bullshitting you the math makes sense.

Edit: when your computer starts using your HardDrive as memory for processing instead of your RAM it is called "SWAP", if you see SWAP at more than 0 you're in serious trouble.
 
Thanks for the reply dude.

My answers are below.

chanilla said:
As a long time LiquidWeb customer, this is pretty common. What kind of site is it? (yes platform, third party code on top, users, etc...)

It's pretty much standard procedure for them to say have your developer mod code, blah blah... implement a request limit at some point, so consider yourself lucky you have the opportunity to fix this... instead of waking up one day wondering why your traffic is cut in half, only to realize your box has been throttled to only allow so many connections per second. (this is usually a problem for any site running a db, without aggressive caching)

Doesn't make them a bad hosting company, and I still have a server there, but it's like going to Best Buy looking for a custom gaming setup... you're at a big box store, wanting services/products outside of the box at times. The trouble is they cater to a WIDE variety of customers, so they're configurations (and support staff) are calibrated for that. Typically they will recommend that you need to upgrade, until suddenly you find yourself on a $600/mo dedi, and are having the same fucking problem as day one. The tier 1 techs are usually the culprit here, and you have to know the right questions to ask before you get anywhere meaningful.

Step 1) site details (type, code, db?, ad code, third party scripts, waterfall screenshot, and how many pages indexed: you want to make sure you don't have a crawler issue if this is a really large site, and it's not actually user sessions bringing your instance to its knees)

WP site, with 1300+ long image heavy posts, 1400 pages indexed.
There are multiple ad codes on the site - probably too many.

Step 2) cache or no cache

W3 Cache installed and hooked up with MaxCDN

Step 3) optimize

Not sure how to do that... waay beyond my skillset.

Step 4) upgrade, migrate, or not
Click to expand...

Would upgrading help? I'm paying about $300 a month and definitely will pay more if it would help.
 
Yup agree with that above. I have a site of similar traffic, can sometimes hit 300 - 400 people on the site at one time.. The mysql requests soon add up!..

I manage it with a decent dedicated server and using rackspace CDN for image files/statics etc... oh a plenty of database repairs/optimizations!
 
MrMedia said:
Would upgrading help? I'm paying about $300 a month and definitely will pay more if it would help.
Click to expand...
Not sure if you missed my answer, but YES, but also WTF? $300 for 3.3GB of RAM?!??! There is no way. In comparison Linode gives you 48GB of RAM for $320 with 768 GB SSD and 12 cores.
 
Just seen your reply thank you Ccarter, the car analogy worked for my brain!

So from what everyone is saying I need an upgrade, pronto!

Would something like this work?
  • Dedicated Bundle
  • Intel Xeon E5-1650 v4
    6 Cores @ 3.6 Ghz
  • 32 GB RAM
  • 2 x 250 GB SSD Primary Drive
  • 1 TB SATA Backup Drive
  • 5 TB Bandwidth
 
Yeah agree. You will get a pretty decent dedi for $300 a month. Not sure if you need a managed service or not though, which will bump price up. Check out reliablesite.net .. excellent network speeds and competitive pricing.

6 core server..
12 x 3.60 GHz
128GB DDR4
1TB SSD+4TB HD
10TB @ 1 Gbps
$229/mo
 
How the fuck are you paying $300 a month for a 4 gig vps instance at liquidweb? You should be on a dedi, but I hate to say it, if you're on Wordpress, you're not going to get the kind of performance you need there in the long run. (ask me how I know!!)

You might look into Kinsta or Pantheon for a more wordpress oriented specialized solution. Cloudways, Linode, Digital Ocean too, but sounds like you need an admin.

However, I think what you're missing is Varnish, Opcache, Redis, Nginx, etc... or a combination thereof. In my experience, W3 and Maxcdn is not enough.

Also, how slow is your site? What does you waterfall look like, and are there any scripts that are producing situations similar to concurrent connections on page load?

If you want to stay at Liquidweb, I would move to a Storm or bare metal dedi. It's gonna be clunky because of how they setup cpanel, but you should get more bang for your buck and be able to have someone optimize the entire server for your needs.

If it were me, I would get the site on Railgun and ditch the other stuff. And contrary to popular belief, I think Varnish would help as well. Check theme and plugins (and I keep saying it over and over... usually people are running third party scripts creating a bottleneck without knowing it) and streamline if possible. Then address your horsepower (RAM). (mind you, all this needs to happen immediately. Every second LW throttles your traffic, or your site is crippled... you may as well start pouring gasoline on your money.)

Alternatively, Joe at Unixy is amazing with custom load balanced solutions, and will smoke Liquidweb any day of the week with a mirrored/load balanced dual server setup... which would be helpful if you plan on scaling up. You may also be able to get a custom solution from Chris at Lightning Base for the same price you're paying now. Either way, both will unquestionably offer better performance than you're currently seeing.
 
Thanks Guys,

I was paying that amount because it included data backup too.

I went ahead and ordered a dedi on Liquidweb and they are building it now.

I got this since it has a lot more RAM than recommended in the thread.

The site is slow but that is because of the excessive ads I have on site. Its a trade off because I need the revenue from each to pay for more growth (content, links, etc) so I accept that it will never be blazing fast.

My main issue was the constant downtimes under a relatively low load. However as CCarter said the 20 people in an 8 people car scenario was my main issue.

Hopefull this should help -

Single Processor
Intel® Xeon® E5-1270 v5
  • 4 Cores @ 3.6 GHz / 4.0 GHz Turbo
  • 32 GB DDR4 RAM
  • 2 x 250 GB SSD HW RAID 1 Primary Drive
  • 1 TB SATA Backup Drive
  • FREE cPanel/WHM OR Plesk Onyx Web Admin Edition!
  • FREE 500 GB Guardian Continuous Backup!
  • FREE Advanced DDoS Protection!
  • $349 p/m

They are going to build it and migrate the site over once done so I will wait eagerly to see what happens when we upgrade...
 
Beyond simply throwing more power and money at it, you can do Front-End Optimization, which can bring in significant savings. But it can only take you so far.

I'd also look into the CMS you're using and investigate optimizing it as well.

For instance, there's a little thing you can do on Wordpress that can drop the CPU usage significantly by telling it to stop checking cron jobs on every-single- frickin' page load. In your case, how many times is it checking? You can tell it to check once per 15 mins, etc.

There are lots of these kind of wasteful processes on CMS's that very few people run into because they don't deal with lots of traffic.

Assuming you hit your optimization peak, you start looking at sharding and load balancing, as well as get onto a dedicated server worthy of your site (I see you may have settled on one.)
 
easyengine.io is the easiest way to get nginx and redis cache running for WP. highly recommended.
 
We run a couple of dedicates with LiquidWeb. When comparing LiquidWeb to other providers they are more expensive, but the customer support is top shelf. To me, extra money is worth it.
 
mate the ads loading are not your site... it's those shitty native ad servers,
 
You must log in or register to reply here.
Back