35 replies to this topic

[ShadowLink64]

Hi everyone,

As you have no doubt noticed, we've been getting a lot of performance-related "500 Internal Server" errors lately. Because it was especially bad recently, we were able to narrow down the issue to MySQL performance problems. Basically, Neocodex's website and database are hosted on two separate machines, and initially we figured that the problem was with the CPU usage on the server our website is hosted on. However, we now know that it is not the website server, but rather, the database server that was causing the problems.

I opened a support ticket with DreamHost asking about the performance issues we were experiencing with our SQL server, and they replied saying that the sysadmin was working on it earlier today, and that they booted a number of users that were abusing the server. As a result, things seem much faster now.

Hopefully we don't see too many more errors, and if we do, at least we know what to attribute the problem to and we can do something about it.

[Waser Lave]

I haven't had any 500 server errors since they did that so up to now it's looking good.

[Noitidart]

YES it is MUCH faster! The search extension I made for firefox, chrome, ie used to not give the live suggestions because it would lag and timeout the request. Now it's working great! Tell the host it has been like this for a LONG time and we want money back especially as this was not our fault.

[ShadowLink64]

YES it is MUCH faster! The search extension I made for firefox, chrome, ie used to not give the live suggestions because it would lag and timeout the request. Now it's working great! Tell the host it has been like this for a LONG time and we want money back especially as this was not our fault.

Nah, I don't want to piss them off. They could very well monitor us and nail us if we run intensive processes, so I want to stay on their good side.

[Noitidart]

Aw ok so the site loads fast but then after it loads the browser says its still loading stuff. But while its still loading everything on the page is usable.

[ShadowLink64]

Aw ok so the site loads fast but then after it loads the browser says its still loading stuff. But while its still loading everything on the page is usable.

It's fine for me.

[Warriors]

Congrats...It's super fast now and no 500 errors yet...Awesome!

[Waser Lave]

It's fine for me.

Fine for me too.

[Warlord.]

At least put LOLcats up next time

It's nice n' speedy now.

[jaredennisclark]

I've been browsing all day and I've yet to get a 500 error.

The site also seems to be responding a lot faster now as well.



Good job. Nice focus.

[Boggart]

I've gotten 2 today

[artificial]

How big is the database? MySQL is only efficient up to a certain point. Maybe you should consider spreading the database over multiple servers.

[aneoguy]

I started getting this alot. but it has seemed to stop now

[Sweeney]

I'm still getting them occasionally, but certainly far less than yesterday.

[iargue]

How big is the database? MySQL is only efficient up to a certain point. Maybe you should consider spreading the database over multiple servers.

We dont have enough users to consider Clustering/NLB.

What we need is MySQL to actually be on the same server as our server. Aka. Lets get a VPS.

But, it will be up to the admins. Hydrogen really wants to move to Django

[artificial]

We dont have enough users to consider Clustering/NLB.

What we need is MySQL to actually be on the same server as our server. Aka. Lets get a VPS.

But, it will be up to the admins. Hydrogen really wants to move to Django

The performance of the database decreases depending on the amount of data and storage engine you're using. You don't need thousands of users to send it in to a meltdown. This forum has over a million posts, a large item database, and who knows what other data (nid?). As the data is hosted on a separate server, it may be cheaper to segment it in to smaller parts, than to rent a dedi/VPS.

The site isn't big enough to warrant a dedicated server, although a VPS sounds like the best bet. You'll get full control over it too, meaning you want have to wait for sysadmins to constantly get rid of the abusive connections.

If you want a decent price, I can set you up on one.

Edited by Artifeetfetish, 02 December 2010 - 03:39 AM.

[Pyro699]

Cant we just blame it all on Noit and move on?

[iargue]

The performance of the database decreases depending on the amount of data and storage engine you're using. You don't need thousands of users to send it in to a meltdown. This forum has over a million posts, a large item database, and who knows what other data (nid?). As the data is hosted on a separate server, it may be cheaper to segment it in to smaller parts, than to rent a dedi/VPS.

The site isn't big enough to warrant a dedicated server, although a VPS sounds like the best bet. You'll get full control over it too, meaning you want have to wait for sysadmins to constantly get rid of the abusive connections.

If you want a decent price, I can set you up on one.

SQL data servers are designed for much much much larger data loads then ours. Dreamhosts SQL server alone has over several terrabytes of data on it. The problem comes with the load to the server itself. The more people requesting, the less people who get the data. We have several servers at work which have over 10tb of data and handle it perfectly.

You clearly have no understanding of how Clustering/NLB works, or what it does.

When you split a database server over many servers, each server doesn't get a "part" of the data. Because then you would waste a lot of time figuring which server to go to. All servers get the same database, and communicate changes to each other. All it does it split the connections among itself to enable more users to connect at once. Clustering gives you more stability as if one server goes down, you have more. Network Load Balance, allows you to support more users.

Our problem is either a)Network lag slows the request. Thats the usually 500 errors, or b)Sql has to many requests, which is why yesterday we where getting errors every second. Hence why SL sent the message, and they said they kicked people off the sql server. Not because they had to much data, but because they where requesting way to much.

[Boggart]

I just got like 5 in a row I even got one while trying to open this thread >_<

[Boggart]

and while trying to post that last comment

[Kyle]

Getting some now

[Ali]

Yep, I'm getting them every other page atm.

[iargue]

Whenever the SQL server hits heavy load, you will see constant errors, but it should clear up faster. We hope.

This is out of our hands for now. Just gotta hope dreamhost can make it better, or the admins move to a different setup.

[artificial]

SQL data servers are designed for much much much larger data loads then ours. Dreamhosts SQL server alone has over several terrabytes of data on it. The problem comes with the load to the server itself. The more people requesting, the less people who get the data. We have several servers at work which have over 10tb of data and handle it perfectly.

You clearly have no understanding of how Clustering/NLB works, or what it does.

When you split a database server over many servers, each server doesn't get a "part" of the data. Because then you would waste a lot of time figuring which server to go to. All servers get the same database, and communicate changes to each other. All it does it split the connections among itself to enable more users to connect at once. Clustering gives you more stability as if one server goes down, you have more. Network Load Balance, allows you to support more users.

Our problem is either a)Network lag slows the request. Thats the usually 500 errors, or b)Sql has to many requests, which is why yesterday we where getting errors every second. Hence why SL sent the message, and they said they kicked people off the sql server. Not because they had to much data, but because they where requesting way to much.

I'm not talking about SQL servers in general, but specifically MySQL databases. It's clear you have no understanding. A single MySQL database with 100GB of data using the InnoDB storage engine (which I believe is what Neocodex is using) is going to perform far worse than having that data spread over 4-5 different databases. If you deny that, you're an idiot. I doubt this is the case for Neocodex, as the site is quite small.

When you get to those large sizes, you don't want to duplicate the data over multiple servers - that's just ridiculous. It is far more efficient to break it up over multiple databases (spanning over multiple servers if need be).

[iargue]

I'm not talking about SQL servers in general, but specifically MySQL databases. It's clear you have no understanding. A single MySQL database with 100GB of data using the InnoDB storage engine (which I believe is what Neocodex is using) is going to perform far worse than having that data spread over 4-5 different databases. If you deny that, you're an idiot. I doubt this is the case for Neocodex, as the site is quite small.

When you get to those large sizes, you don't want to duplicate the data over multiple servers - that's just ridiculous. It is far more efficient to break it up over multiple databases (spanning over multiple servers if need be).

Our data is in multiple databases.......