What settings for maximum speed on a large configuration ?

[Monarobase]

Hello,

I've recently installed litespeed on our pre production server and when I get all the issues sorted out I will be transfering quite alot of accounts from a smaller server running Apache to this new server.

I've read in various places that setting some options too high will slow things down instead of speeding things up.

Our customers host various scripts (Wordpress, Joomla, PhpBB, vBulletin, IPB and also custom scripts, some better optimised then others).

We're aiming at getting the best performance and stability for our customers websites.

Our server has got 256GB of RAM and 32 * 3.1GHz threads.

I've already activated and configured Xcache 3.0.1 and allowed it 1GB of cache space…

I've run some tests and seem to be getting 503 error messages when there are too many processes being run. By upping LSAPI Max connections and LSAPI_Children from 35 to 400 most of the errors have gone.

I want this server to be able to cope with a large number of simultaneous requests and serve each of them quickly. I want to try and get the configuration right before migrating the accounts to it.

Can you guide me to have the right settings for maximum performance and maximum number of processes ? Shoud I pay to have you configure these settings ? Can you do this with only a few customers on the server ?

I've been monitoring the performance of a small php script with echo "hello world" from different locations and have been getting spikes every now and again.

Most of the time the scripts download in 15 to 17ms and then every now and again the latency goes over 500ms…

This server has only got a few medium traffic sites on it so it should be stable and have regular latency times. I'm worried that if I get 503 errors and latency spikes what it will be like with 700 or 800 accounts on the server.

The server load is always below 1.5 for the mement, this includes MySQL and other services so I don't think I'm being limited by the 2 Core test licence… am I ?

Thanks

[waski]

I'm also in a position simmilar to Monarobase's. We are planning to migrate to Litespeed with hundreds of accounts, thousands of domains. Please advice how to avoid 503 errors and which parameters are most important in shared enviroment?

Also Monarobase - haven't you consider putting LSPHP to separate machine/s just for redundancy/scalability ?

[Monarobase]

Originally Posted by waski

Also Monarobase - haven't you consider putting LSPHP to separate machine/s just for redundancy/scalability ?

What do you mean exactly by this ? Activate Litespeed loadbalancer ?

Or have a NAS shared by two servers ?

Or something else ?

[waski]

I don't mean http Load Balancer but load balanced external app (lsapi). Of course data gotta be shared with that external app - by NAS for example.
That way - php can be compiled on other machine/s - thus giving more flexibility and scalability - thats what we have in plan.
It's just not a best idea to keep all of that stuff including mysql (and probably mail?) on same machine while you mentioned 800 accounts.

[Monarobase]

We would have to test the speed to see what the extra network would add…

I'm not sure about a reliability difference as if you don't also have the web server then the PHP scripts alone won't be and it seems that you would be adding another point of failure unless you can automaticaly share all PHP processes between both servers.

The first thing we would move would by MySQL to create a multiple server cluster. We've alread clustered DNS servers but that doesn't use much ressources.

The server we're currently trying to get running smoothly with litespeed has more than 5 times CPU than our previous server and 10 times more memory so with the right software and settings we should be able to hold out quite some time on the new server configuration.

If you're using a NAS for the files, why not go the whole way with a load balancer ?

[waski]

unless you can automaticaly share all PHP processes between both servers.

Not quite following.

Originally Posted by Monarobase

If you're using a NAS for the files, why not go the whole way with a load balancer ?

As far as httpd is easy to load-balace, Direct Admin which we use - is not :/
Hence it's easier to move hard-computations (php) to separate servers while main httpd server is not that busy.

We are getting off-topic a bit, hope someone will solve your 503's Have you tried to turn xcache off, and how are your hdds doin?

[webizen]

Originally Posted by waski

I'm also in a position simmilar to Monarobase's. We are planning to migrate to Litespeed with hundreds of accounts, thousands of domains. Please advice how to avoid 503 errors and which parameters are most important in shared enviroment?

...

https://store.litespeedtech.com/stor...yarticle&id=31

[Monarobase]

Hello,

I've read the article, I'm getting 503 error messages and have actived core dumps but am not getting core dumps when I get the 503 error message.

I've checked my logs and it seems to happen when /scripts/restartsrv_httpd is run.

I'm watching to see if this is the problem, if it is I will ask you for some advice about making this script do a gracefull restart of litespeed instead of completly restarting litespeed and all php children.

cPanel runs this script by default for each account when it rotates logs so it's important that litespeed is reloaded without causing 503 errors.


Here is the type of errors I'm seeing :

2013-01-22 08:22:42.168 [INFO] [XX.XX.XX.XX:51376-1#APVH_my_domain.com] connection to [uds://tmp/lshttpd/lsphp5.sock.687671] on request #0, confirmed, 1, associated process: -1, running: 0, error: Connection reset by peer!
2013-01-22 08:22:42.175 [INFO] [XX.XX.XX.XX:51376-1#APVH_my_domain.com] connection to [uds://tmp/lshttpd/lsphp5.sock.687671] on request #0, confirmed, 1, associated process: -1, running: 0, error: Connection reset by peer!
2013-01-22 08:22:42.183 [INFO] [XX.XX.XX.XX:51376-1#APVH_my_domain.com] connection to [uds://tmp/lshttpd/lsphp5.sock.687671] on request #0, confirmed, 1, associated process: -1, running: 0, error: Connection reset by peer!
2013-01-22 08:22:42.183 [NOTICE] [XX.XX.XX.XX:51376-1#APVH_my_domain.com] Max retries has been reached, 503!
2013-01-22 08:22:42.183 [NOTICE] [XX.XX.XX.XX:51376-1#APVH_my_domain.com] oops! 503 Service Unavailable

[Monarobase]

Ok I've been watching the server logs all morining. I've had multiple error 503's and everytime a second or too before I have something like this :

2013-01-22 12:07:04.347 [NOTICE] Server Restart Request via Signal...
2013-01-22 12:07:04.347 [NOTICE] [PID: 777249]: start cgid: 878637, /usr/local/lsws/bin/lscgid
2013-01-22 12:07:04.347 [NOTICE] [Child: 777254] Start shutting down gracefully ...
2013-01-22 12:07:04.347 [NOTICE] [Child: 777255] Start shutting down gracefully ...

So I'm glad it's not actually something crashing but I cannot have 503 erros on my customers websites everytime cPanel requests a reload for Apache or Litespeed is gracefully restarted.

What can I do to stop this from happening ?

[NiteWave]

graceful restart will cause 503 error --- this is a good point, and new to me. usually 503 error is random and not easy to reproduce. not sure if you can work out a simple test case which we can reproduce it in our lab so we can proceed.