I am working to piece together a Debian Squeeze Perfect Server ISPConfig 3 install that meets all of my needs. For this install I'm using the following guides: Perfect Server ISPConfig 3 Debian Squeeze by Falko Timme Perfect Server, Multi-server ISPConfig 3 Debian Squeeze by Till Brehm Perfect Server, Multi-server ISPConfig 3 Debian Squeeze and OpenVZ by Michel Kaiser What I would like to know is, if I install Quota as specified in Falko's document on step 13, is that going to cause problems for me when I try to create my virtual servers? I'm not sure what to do here. I have had problems in the past trying to install OpenVZ after installing quota. My base system is going to have Mail, Database, and Web, then I will have two virtual name servers NS1 and NS2, then I'll have one virtual git server that I use for my dev repository. Please help give me guidance on the best way to approach this install keeping in mind that I would like to be able to monitor user quotas. What is the best way to have user quota monitoring and virtual servers using OpenVZ without causing conflicts? Thank you, Darin
Ok, I've been working on my server setup today, and the answer is no as far as I can tell. OpenVZ would not install a virtual server with quota on, so I turned it off and it worked fine. My primary server is not virtual, so how do I use quotas with it? Do I need quotas, or will ISPConfig 3 handle it for me?
Why not install Web/DB & Mail in OpenVZ containers as well? That way you'll have perfect isolation, easy backup/migration etc. With OpenVZ's new ploop layout i/o Benchmark is pretty well, so you won't have any bad performance impact. Tricky part is then quota inside container -> see here: http://www.howtoforge.com/forums/showpost.php?p=292363&postcount=2
My thoughts on VM Hi MaddinXx, That's cool stuff. Nice contribution. How is the script envoked, when performing certain activities within ISPConfig 3? Does it automatically take care of setting quotas via settings in ISPConfig 3? I read the OpenVZ page that you provided this one when I was doing my research on OpenVZ, but I'm very new to OpenVZ, so I have no idea what performing the install that way would do to my system. Sounded like maybe the kernel I would be stuck with after installing would be Redhat, so I didn't even try it. Not to mention that I don't understand what ploop is or does. I tried several OpenVZ installs without success: The first one I tried was this one using OpenVZ Web Panel. It worked after I reinstalled the server without quota. After that install, I wasn't satisfied, because I could not access the Web Panel under https, so I tried these: Installing And Using OpenVZ On Debian Squeeze (AMD64) Installing OpenVZ + Management Of VMs Through ISPConfig 3 (Debian 6.0) I would prefer not to run an additional web server on the base system, if I can get that type of install to work. It has been painful trying to get it to work though. I read tons of information trying to get the later two installs working, but never succeeded. This is a pain, because I didn't want to run another web server on the system just for installing OpenVZ under the Web Panel, but I had no choice since I couldn't get any of the other installs to work. I even went as far as trying to enter everything for the server at the command line using this page, but that didn't work either. Finally I found a way to start the web panel server under https, so I decided to go with what worked: https://code.google.com/p/ovz-web-panel/issues/detail?id=228 I was and am concerned about performance installing the main system components under a virtual server. The server that I'll be using has dual quad core Xenon 2.33G processors, 16GB RAM, 2.7T RAID 10 disk array, but I want the best performance that I can provide my clients and I thought that would be doing the install on the base system, and only using virtual servers as resolvers and my git repo. I don't really care for the multi-server setups due to all of the resource sharing as well as the added work to keep everything up-to-date. I want to minimize the resource load on my server, have one single database to backup, one web server to manage, which means I don't have to have an additional web server for the mail server, which is a pain when it comes to managing the certificates and making everything work. What are you backing up when you backup your servers? Are you backing up the entire virtual containers? If yes, I don't understand how that is easier than dumping one database. Can you explain your thoughts on this? You are much more experience than I am at this, so I would like to hear your thoughts about my approach... Is the install provided here a sure thing, or do you think I will run into a lot of problems? Thank you, Darin
It runs on the HW node, not within the virtual containers. You do not have to do anything special within the VE, just use ISPConfig/quota as usual. The OpenVZ quota system will take care of the rest. It's safe to go that way - and like the easiest to get current versions of the kernel/tools (which include swap support, ploop etc.). Of course you'll have "RHEL based" kernel afterwards, but you won't notice that. PS: ploop is a new format/layout like: 1 container = 1 file. A lot of virtualization plattforms tend to go that way since it will reduce disk i/o problems. OVZ Web Panel uses built-in Webrick server - consumes nearly nothing. Make sure to edit the file /etc/owp.conf to enable SSL (config within OVZ dir will be ignored) You will not get the same performance within virtualized servers as with "native" ones (what I guess, have read etc.) - and if you really don't see great benefits for your backup strategy etc. then I'd not go with containers and use the HW node. But I'd think once more about it... There are great ways to centralized manage your servers (like Rexify framework) that allows you to fire 1 command that gets executed on all connected servers (e.g. you upgrade all servers at once). To summarize, it's definitely up to you. Me however, I will never leave my VEs anymore. I already had enough situations where I was more than happy to be able to restore the whole server with just a few clicks - or migrate to another server with more power because the current one wasn't strong enough anymore. For backup, I primary backup the VEs on a daily base. If you ever need to restore some single files, I unpack the backup of create a temporary virtual container where I can do all backup jobs as usual (e.g. MySQL dump). but again, it's true - if you have software (like DBs) that have millions of queries and you already encountered boatnecks, a VE might not be the best solution. But as long as you have "common" hosting, this won't be the case. And to get best/better performance, it's more a question of good configurations (apache, nginx etc.) rather than: Native vs. virtualized. Hope this helps. Regards
Great feedback! Thank you... That's great! I'll take a closer look. Yeah, you're right, i'll be working in the VMs, so who cares what the base node is as long as it's secure. I've been using this to start the secure web server, and that allows me to use OpenVZ Web Panel securely: Code: ruby /opt/ovz-web-panel/script/server_ssl webrick -e production -p 3001 -b 0.0.0.0 -d I'll take a look and see if I can figure out how to set the configuration to use the SSL port. Are you running a web server on your primary node to manage your VMs then? I'll look into reflexify framework. Thanks... Are you using ISPConfig 3 to backup your VMs? I just purchased a Dell PowerEdge 2950 III, 2U system that has 2 2.33 Intel Xenon processors, 16GB RAM, and has 6 450GB SAS 15K.6 drives in a RAID 10 configuration. This server is being configured now, and will go into a local data center on the 15th of March. I think I can probably get away with running a few VMs as I am doing now. Thanks for taking the time to reply back, and for the valuable information! Regards, Darin
