Problems With Gamma Server - MYSQL service is bouncing connections

Hey guys there seems to be a problem with teh MYSQL service on Gamma..ITs giving a "Too Many Connections" error message to any request its being given at the moment..also I notice that the trap17 hosted server seems to have the same errors..is that gamma server too? On the server status page its listing as the MYSQL service as up and running and ok ..but its not.. I think it needs a reboot..hopefully this can be sorted soon.. Thanks!!

Comment/Reply (w/o sign-up)

Yes, the Trap17 and AstaHost hosting accounts are on the Gamma server. The MySQL error is a normal part of the servers operation. The error message is given in one particular situtation. It is triggered when there happen to be too many connections to MySQL at the same time! As I have explained in other topics similar to this one, the MySQL server has a limit on the number of simultaneous connections to it.

Lets say that the limit is 100 simultaneous connections! That means that if there are 101+ connections requested at the exact same moment, the error appears! Since most connections only last hundredths of a second, this allows for thousands of connections per second. So it really takes a lot of people on the server at the same time to cause the problem. Even if you had 3000 people viewing material on the Gamma server, what are the chances that all of those pages even required a MySQL connection and if people are reading content, what are the chances that they all would choose to load a page at the same time?

My point is that several thousand pages can be viewed at the same time before the server notices any problems.

Now lets consider other variables in the equasion. I'm guessing that you posted this article at the time of the outage. It happens that the time that you posted this topic is also one of the busiest times on the internet each day.

So we got a whole lot of people viewing our pages all at once and then we got one too many MySQL requests at the same instant and bang!!!!!!!!!!!!!! The error message starts popping up for everyone. So what is their first instinct?

"Click Reload!"

So now the server is kind of overwhelmed and then a couple of thousand people click reload at the same time. Now the server is completely overwhelmed. It can't even process all of the error messages as fast as the errors pop up resulting in connection request lasting longer and using up the time that could have been used for an actual connection. So we get another reload request which then makes things even worse. etc...

Maybe the MySQL server needs some kind of queuing system but I don't know of one. The best possible solution is for software developers to write their code to respond differently to the "Too Many Connection" error. If the script forced the user to wait 30 seconds before displaying the message, then the server would have time to recover before a barrage of page reloads.

Basically, once this happens, the problem usually grows out of control until enough users are discouraged and break the cycle by not reloading.

It would be nice if you could add code to your robot.txt file that would allow you to specify the time of day that it is okay to crawl your website which could at least reduce the number of connection requests by eliminating the bots during peak hours. Some web software allows a website to be dynamically generated once a day and then the visitor actuall sees a static page. This doesn't really work for many dynamic websites like a forum or blob but for a link index or daily news site, it would be fine.

Really, all we need is for users to analyze the situation wait a minute and then reload.

Long story summarized this happen very infrequently and is only a problem during peak internet hours. Not only is this problem compounded by users not reacting to the problem in the logical manner, the way that scripts handle the error and the number of connections a script requires to load a page are also major issues. The server produces the error message so that it doesn't need to be restarted. Basically, there isn't much that can be done to resolve the issue when it arises.

Hope this helps.

vujsa

 

 

 

Comment/Reply (w/o sign-up)

Sure I understand all of this being a server admin myself for many years at many different companies..
Just find it odd that the peak time here on this server is around 10-11pm euro time..normally earlier in the day for this type of hammering!

Btw any news on the CRON JOBS? we REALLY need that problem fixing soon!!
The cron service seems to have gone wrong during the cpanel upgrade.

Thanks!

Comment/Reply (w/o sign-up)