Ngemu problems

[PentiumPro]

Whats going on with nGemu right now? Are you guys trying to fix something?

[Betamax]

Moved to feedback.

[Kirby]

Alot of theories are going around:

1) The MyDoom virus that is going around might of struck
2) An upgrade went seriously wrong
3) your probably just seeing things

[tonyyeb]

II tthhiinnkk ii''mm sseeeeiinngg ddoouubbllee..... or not... not seeing things, just must be problem due to the popularity of ngemu.com

[Keith]

Almost everyday at some point it is the same error on the mainpage. Not sure the cause either, but just give Bobbi time and he will get it fixed. He is only one person and doesn't always have a lot of free time to look into this stuff and keep up on it on a regular basis.

[tonyyeb]

Thats true Keith. I just hope he never looses interest in the site cause he puts a lot of hard work in to keep it running and I know from talking to Dixon that it's and expensive thing to keep running.

[tonyyeb]

Well still no joy on getting on the front page, too many connections... needs a yank of the chain (aka a good flush!)

[Cypherswipe]

Warning: mysql_connect(): Host 'srv01.ngemu.com' is blocked because of many connection errors. Unblock with 'mysqladmin flush-hosts' in /home/virtual/site10/fst/var/www/html/modules/functions.php on line 13
There was an error connecting to the NGEmu database. Please try again later.

That sounds like it can only be fixed by an admin, not something that will go away when some of the users log off.

[roxor]

I have been having so much trouble getting into emuforums lately. The damn connection keeps timing out. WTF is going wrong?

[FLaRe85]

Originally Posted by Cypherswipe

That sounds like it can only be fixed by an admin, not something that will go away when some of the users log off.

Correct. It's a problem with the way the MySQL server is configured. The only two ways to deal with this are flushing the hosts file and setting the error connection counter to a higher value or even unlimited. Bobbi has been meaning to fix this, but hasn't gotten around to it yet.

[Velocity 7]

Well, I hope he gets to it now, because things not fixed early tend to haunt people later.

[Sniper Wolf]

Umm is it me but all the graphics on the main page isnt there its going strange

[FLaRe85]

Originally Posted by Velocity 7

Well, I hope he gets to it now, because things not fixed early tend to haunt people later.

Bobbi does what he can. Nobody pays him to run this site. He does this as a hobby and pays a tremendous amount of money so that you can all have NGEmu. I think it would be wise to respect that he can only fix things when his schedule permits him to do so.

[Cypherswipe]

True, it's just aggravating when these problems last for a day or more.

In any case, the site seems to be working agin now.

[[Unknown]]

I would assume this site is hosted on a dedicated server, running Apache... or rather, I know such because this site sends these headers:

Server: Apache
X-Powered-By: PHP/4.3.3

As such, I recommend messing with whether the database uses persistent connections or not. Basically, if this server is using them, it might be creating too many children and causing errors. If it's not using them, it may be that too many people are connecting at once.

If you have administrative (root) access to the MySQL server, there's a lot more you can do. Mostly, these queries:
SET GLOBAL max_connections = 500;
SET GLOBAL max_user_connections = 0;

There, of course, is a lot that can be done to make this server faster; caching, for one thing - but I seem to remember seeing stuff like that was being done.

-[Unknown]

[Bobbi]

Originally Posted by [Unknown]

I would assume this site is hosted on a dedicated server, running Apache... or rather, I know such because this site sends these headers:

Server: Apache
X-Powered-By: PHP/4.3.3

As such, I recommend messing with whether the database uses persistent connections or not. Basically, if this server is using them, it might be creating too many children and causing errors. If it's not using them, it may be that too many people are connecting at once.

If you have administrative (root) access to the MySQL server, there's a lot more you can do. Mostly, these queries:
SET GLOBAL max_connections = 500;
SET GLOBAL max_user_connections = 0;

There, of course, is a lot that can be done to make this server faster; caching, for one thing - but I seem to remember seeing stuff like that was being done.

-[Unknown]

We're not using persistent connections, that'd be crazy. The settings you suggest are good, however already being used. We're using mySQL caching as well as a self-made caching system, which however is far from being perfect (but reduced load by 40-60%, depending on the number of people being online)

That the errors happened are related to +10.000 errors having happened within a fairly short amount of time ... we are looking into finding a way to get less errors, but I fear this won't happen before the holidays (22th February), as I'm simply too short on time.

Thanks for sticking with us, we're working as much as possible to get these issues worked out.

[tonyyeb]

Big up the support for Bobbi and the crew... or something like that! Keep up the hard work!

[FLaRe85]

Originally Posted by Bobbi

We're not using persistent connections, that'd be crazy. The settings you suggest are good, however already being used. We're using mySQL caching as well as a self-made caching system, which however is far from being perfect (but reduced load by 40-60%, depending on the number of people being online)

That the errors happened are related to +10.000 errors having happened within a fairly short amount of time ... we are looking into finding a way to get less errors, but I fear this won't happen before the holidays (22th February), as I'm simply too short on time.

Thanks for sticking with us, we're working as much as possible to get these issues worked out.

Right. I don't think people realize what is going on here. So, here's a more technical definition:

We have 2 dedicated servers. When you connect to www.ngemu.com, you are connecting to the main page, which is hosted on srv01. However, the MySQL server is hosted on srv02. So, in order to dynamically build the front page, index.php causes srv01 to make a connection to the MySQL daemon on srv02 in order to run the SQL queries. This is why the message indicates that srv01.ngemu.com is blocked because of too many connection errors. What this error means is that srv01 is temporarily being banned from connecting to srv02 after it reaches a certain amount of "broken" connections defined in an error counter variable. It does this is actually for security reasons.

Why are so many connections being broken so quickly, I have no idea. That could be a network issue, but it's hard saying. So, for the time being, the only way to make sure this doesn't interfere with the operation of the site is to raise the error counter variable.

[[Unknown]]

ShADoWFLaRe85, I've seen two different messages. I've seen the one you describe (it was up for days...) and another - that there were too many connections to the MySQL server from the same user. This is why I suggested tweaking the max_connections variables.

You might also look into using a slave/master configuration. Using vBulletin's current system, and likely the one you use on the main site, you could send different queries (write vs. read) to different servers.

The reason for this is simple. If you have srv01 running MySQL, but read only, then it can do all the data for the main site. No matter how far apart the ping is for the two servers, (by the by, they're in different areas, which is very slow between servers.) having a localhost/pipe connection is faster. So, if the master (srv02) is sending all writes, parsed, to the slave (srv01) it could speed things up considerably. (depending on how your caching system works, and how much is caches.)

http://www.mysql.com/doc/en/Replication.html

Bobbi, you might check the queries in vBulletin and your page to make sure they are largely static. Query caching, as I'm sure you know, works off the query itself staying the same - so if it is dynamic with WHERE conditions that depend on this page view, or with time() in it, this will cause problems. If these can be changed at all, it would improve things a lot.

However, this may or may not solve the issue. There are other things, of course, that can be done. I'm more than willing to help with this, if you're willing to accept my help. I'm not trying to tell you things you already know or patronize you, just trying to give you ideas that you might not have thought of because they were so simple.

And, persistant connections, yes, are a very bad idea considering the distance between your two servers.

If you want to know who I am, I'm the lead developer for some as-yet unreleased forum software, SMF. I do a lot to make sure it's fast and to optimize it's queries almost every day, so I might actually know something about what I'm talking about .

-[Unknown]

[mikeshoup]

Hey [Unkown]! Just want to say, good job on SMF :P I'm using it in an upcoming site of mine. Thanks for that wonderful work you do!