Fuckloads of updates

Fuckloads of updates

This post in a nutshell:

• Far less lag on the servers.[/*:m:2zyvz3t3]
• Very good chance Remote Desktop will no longer affect server performance. [/*:m:2zyvz3t3]
• Good chance that no more full dedicated server reboots will be needed for quite a while (i.e. no more random MySQL crashes).[/*:m:2zyvz3t3]
• Faster map changes (less hard disk usage).[/*:m:2zyvz3t3]
• Faster high-load websites (shadiku.com, gamingmasters.org, reupper.com, imgkk.com etc.)[/*:m:2zyvz3t3]
• No need to hide SRCDS windows every time you log into RDP.[/*:m:2zyvz3t3]

------------------------

This posted started out as a huge rant about Perma Run, which I have now decided against using for various reasons. There's still a lot of interesting changes besides Perma Run though


*snip huge block about perma run*


[strike:2zyvz3t3]During my latest coding spree, I have created a script known as Perma Run, which will execute any given file asyncronously on the SYSTEM account, much like how SMR executes the SRCDS servers.[/strike:2zyvz3t3]

[strike:2zyvz3t3]This will allow us to have the main Session 0 account completely logged off at all times, resulting in lower RAM usage and better performance for processes running on Normal priority (Apache, MySQL etc.)[/strike:2zyvz3t3]

[strike:2zyvz3t3]However, in doing so, we have essentially removed the ability for any alternative for RDP (all of which were rubbish) to be used - only an Remote Desktop connection will log in and start the essential background apps, such as explorer.exe itself.[/strike:2zyvz3t3]


I have also discovered that, for whatever reason, the SRCDS servers were being started at Normal priority, which is the sole reason for lag while someone was accessing RDP or the MySQL server was getting hammered by HLstatsX, which happens more or less 24/7.


Also, I have discovered a (but probably not the only) cause for the strange MySQL crashes - Diskeeper is defragmenting the binary logs and similar InnoDB-related files while MySQL is trying to access them, therefore crashing it. MySQL is very lenient about how it opens handles to files, so Diskeeper doesn't know any better - normally it would ignore locked files. I have added MySQL to the Diskeeper blacklist - let's hope this means the end of MySQL crashes (which inevitably cause the whole dedicated server to need a full reboot). Unfortunately, MySQL will eventually get pretty slow from fragmented files (I really hate NTFS sometimes...), but there isn't much I can do about it other than the occaisional manual defrag with MySQL not running.

I have also found and fixed a couple of errors which caused a 4gb log file to build up, lol.

Also, for some reason windows was set to allocate CPU to background services (WRONG when primarily running SRCDS servers), and had a fixed virtual memory size, which Diskeeper wasn't too happy about. Both these problems have now been fixed.

Also, SRCDS servers will now be automatically hidden with 5 seconds of starting - less SRCDS spam after reboots.

And please stop using derogatory terms such as RDP-monkey, or I will take action against voi.

Anyway enjoy the updates

 

Re: Fuckloads of updates

Oh ok I thought KJ was in voi. Bun definitely is though, right?

It was writing continuously to the Apache log file and PHP error log, which I have now disabled/fixed respectively

The forums are getting pretty active

 

Re: Fuckloads of updates

RDP only made the servers lag because Windows prioritised a smooth RDP experience over SRCDS performance D: