Settings for LVE default suggestions and opinions please?

[scott]

First CL server, setting up shared hosting packages.

[dave518]

Hi CL forum folk,

Im really getting near putting my first CL (6.7) server into production and preparing to migrate users fr om old RHEL servers to it, and I could really use everyones opinions and advice on Settings for LVE defaults for the hosting packages I provide on my shared hosting biz.

Of course Im talking about these:
CPU usage (SPEED)
Virtual Memory (vMEM)
Physical memory (pMEM)
Concurrent connections (EP)
Number of processes (nPROC)
I/O lim it (IO)

I know that those settings should depend on how the server is being used and the types of hosting clients, content & software running, and kind of data usage, so Ill describe my users / environment in detail here in hopes that folks here are willing to provide your opinions and advice.

I basically have 4 "levels" of hosting packages I offer and they go something like this:

Lev 1 - Fairly basic users, small amount of data (under 1GB), a single WordPress, and few email addresses.

Lev 2 - Users with up to 5GB data stored, up to 3 WordPress (or similar database-driven scripts), 10 email accounts and fairly active blogs.

Lev 3 - Users with up to 20GB data stored, mix of multiple PHP scripts and 6 mySQL databases with heavy use on images and e-commerce, lots of uploading via PHP scripts, 100,000 files.

Lev 4 - Users with 40GB data stored, mostly large image galleries in multiple scripts like WordPress, OpenCart, Forums, e-commerce with SSL and dedicated IPs. Very high volume of image file uploads and in some cases hi-res images for print, transferring over 100GB bandwidth monthly. Lots of user sessions in database and tracking / cart functions. Gallery scripts like Coppermine and Menalto Gallery processing lots of simultaneous image uploads and lots of mySQL constant activity.

Almost all customers are small businesses who depend on me to provide good service. Not many casual users with hobby sites.

The Lev 3 and Lev 4 users are doing high-volume emails. Not so much bulk mailing, and definitely no outgoing spam, but because theyre businesses and publish their email addresses publicly on places like Facebook they get a lot of spam for Mailscanner to have to process.

Each of the plans is priced different obviously ranging from extremely cheap to somewhat pricey for the high plan.

Each level has its own package in WHM.

My hope is to get your advice on what 4 sets of LVE settings numbers you would use for each one.

In case it matters I also have all the ConfigServer scripts installed - CSF/LFD and CSX and Mailscanner etc...

The server specs are - Dual E5-2620v3 CPU, 32GB RAM, and 2x 1TB Solid State Drives. Second drive is mounted as backup and cPanel backup runs nightly backing up all user accounts / DBs / system files from the first drive to the second drive, and the server will be at least 600GB full so there is quite a lot if I/O going on.

I know Im asking you to spend some time to post a reply and I really truly appreciate any kinds of responses, no matter how little or how much guidance you feel like providing.

Even if you just post this with numbers next to it would be a big help:
CPU usage (SPEED)
Virtual Memory (vMEM)
Physical memory (pMEM)
Concurrent connections (EP)
Number of processes (nPROC)
I/O lim it (IO)

Great thanks for your advice and experience shared, I really need the input!

[sqh]

I had these same questions when I first started using CloudLinux, and found it very frustrating that there were no guides on helping us to make informed decisions on how best to set these levels. In the end, we had to just use the defaults, and then see how it worked, and tweak from there.

Im glad to share with you some common settings for our environment (somewhat similar to yours)

CPU usage (SPEED) 200% (most of our servers have 8 cores, so this lets a user burst to 2 cores @ 100% each)
Virtual Memory (vMEM) 5120MB
Physical memory (pMEM) 1024MB
Concurrent connections (EP) 50
Number of processes (nPROC) 100
I/O lim it (IO) 4096

Hope this helps!

I think the other very frustrating thing you will find is that, despite an incredible amount of complaints, CloudLinux still does not properly alert you, as the server owner, when you have a customer hitting faults. We have put something together ourselves, with some tips from this forum, and its working OK. But its very disappointing that CloudLinux doesnt pay more attention this important aspect of running a CloudLinux server. Your mileage may vary.

- Scott

[dave518]

I thank you for taking the time to respond, Scott. And yes - one of the features that attracted me to CloudLinux away from RHEL was the ability to be alerted about troublesome user accounts. I havent read quite enough here yet to discover what you just told me about it not delivering the alerts that we can enable in the settings. Thats kind of a bummer actually.

But Im still hopeful that switch from RHEL to CL is going to make it easier for me to identify and more importantly throttle a couple of users who are trying to do way too much (like an OpenCart user with 10,000 categories and 200,000 items for sale in it, running campaigns... ouch says the CPU and I/O when that user tinkers with their site each night). Hoping that CL is going to give me the ability to stop that account from hogging up so much and spiking the server load every couple nights, which is currently happening with them.

I do have a question about one of the numbers you provided in your example - Virtual Memory (vMEM) 5120MB - someone recently suggested to me that I should set that to ( 0 - Unlimited ) based on the kind of description Im giving. Did you discover a particular reason why thats a bad idea?

Im definitely curious to know how you came up with Virtual Memory (vMEM) 5120MB

Edit - Im also very curious how you decided on I/O lim it (IO) 4096 (same someone suggested 1520 to me, and as you can see Im very unsure of myself starting out with CL)

And seriously - thanks again very much for your response, it really did / does help!

[bogdan.sh]

I believe Scott did a typo in 5120MB , it should be 512MB. However yes, limiting only by pmem is better way and we recommend setting vmem to 0.

I just replied quite similar topic about better limits, you may want to check it out: http://www.cloudlinux.com/solutions/...m11/topic1081/

There is a possibility to get notification when someone is hitting limits, for cPanel it could be configured from WHM > CloudLinux LVE Manager > Options, with enabling Notify Admin option and setting minumum number of Faults to notify admin = 1 . Like on this image: https://www.dropbox.com/s/2kjq5aq45r...54-15.png?dl=0

[sqh]

Hi Bogdan and Dave.

That was not a typo, we are using 5120MB for vmem. I would be interested to know more about this setting and the recommendations.

Bogdan, while you CAN enable alerts, they are very worthless, as you are probably aware. They will tell you that a client hit a limit, but wont tell you what limit they hit, forcing you to log in and view history and try to reconstruct what happened. We created our own script to read the logs and put together a useful alert like this:

> fubaruser exceeded one or more resource limits w/in the last 1 hour(s):
> - Exceeded 1GB physical memory limit. Site was not available 6 times.
> - Exceeded max disk IO rate. Used 5KB of 5KB. IO speed was slowed.
>
> Processed line: fubaruser,0,8,19,0,26,50,5316723,1308299264,429496 7296,0,0,5527869,1073741824,1073741824,0,27,50,6,0 ,14,5120,5120,1,568,1280

This is a very useful alert, unlike the alerts that CloudLinux sends. Dave will see for himself soon. :-)

P.S. for some reason it shows the io as 5KB but its really 5MB. :-)

- Scott