Portal Home > Knowledgebase > Articles Database > Realistic -- how much downtime to upgrade from cpanel CentOS 5.8 to 6.3?
Realistic -- how much downtime to upgrade from cpanel CentOS 5.8 to 6.3?
| Posted by mrzippy, 09-25-2012, 08:07 AM |
| Hello,
We want to be able to use the CloudLinux disk io limiting feature, which is only available on CentOS if you have v6.x. (We are using cpanel.)
Can anyone tell me what is a realistic downtime expectation for upgrading from CentOS 5.x to 6.x?
Also, what is the recommended upgrade path? Half my admins say we should setup a new server with v6.3 and then migrate all customers to it.. and the other half say we should upgrade on the same server.
Apparently, it's better to "start fresh" rather then upgrade.. but of course the down side would be the inevitable problems that come with moving websites from one server to another. (php config issues, apache issues, the endless minor issues that customer's will find and complain about, new IPs for all resellers, new IPs for any sites that have offsite DNS, etc...)
I would *vastly* prefer to upgrade without changing servers.. but am not sure what are the drawbacks of doing this.
Thanks for any thoughts and/or advice.
|
| Posted by BestServerSupport, 09-25-2012, 08:29 AM |
| You can directly upgrade using yum update. It will not take more than an hour. However, to upgrade like this way is something not recommendable. I would recommend you to first set up a fresh server with new version of CentOS and then manually move websites. You can test websites on new server using temporary URLs.
cPanel/WHM provides excellent transfer feature to transfer websites between cPanel/WHM servers. Please note that if you upgrade on same server and problem arrives then you will be in a big trouble.
|
| Posted by RRWH, 09-25-2012, 08:30 AM |
| A clean install is preferred for major version upgrades - and that would be as you describe - build a new box, configure it and then migrate all your clients onto it. Time - yeah, a few days to build, secure and configure the new box and 2-3 weeks to advise and migrate your clients.
An in-place update should be just a matter of using yum to do an update and then a re-boot. BUT, if something screws up, you will have to restore everything from backup and this could potentially cause major outages while you recover everything and restore all your cpanel backups.
Time for an in-place update would be a LOT shorter if everything goes well, but a lot longer if everything goes south. It is a matter of risk/reward. Personally, with only a small number of servers, a re-build and migrate would be what I would do. If you have a large number of servers then I would probably look at setting up a test server with a handful of real accounts on it (with backups ready to go) and mirror the system, break the mirror (to have a roll-back) do an in-place update while documenting the steps. Once I was confident, then plan the in-place upgrade - including the whole recovery process (break raid1 mirrors).
No matter which way, it is a non-trivial matter to ensure minimal downtime for your clients.
|
| Posted by vpsshell_co_uk, 09-25-2012, 08:35 AM |
| Hi,
Please be aware that upgrades between major versions are not recommended:
http://wiki.centos.org/FAQ/General#h...90e887b1b12a00
|
| Posted by Dan_EZPZ, 09-25-2012, 08:42 AM |
| You can just switch to the CloudLinux hybrid kernel: http://docs.cloudlinux.com/index.html?hybrid_kernel.htm
|
| Posted by NHRoel, 09-25-2012, 09:53 AM |
| As far as I am aware of, Centos doesn't support major upgrades (such as 5 to 6 and so forth). Another reason why I love debian. However, there are ways around. But to get the most out of Centos and avoid further issue, install a fresh minimal version of centos 6 on a new server and then transfer your data over.
--Roel.
|
| Posted by mrzippy, 09-25-2012, 10:00 AM |
| This is a good idea, but I'm having a difficult time trying to understand what are the negatives about using the hybrid kernel.
ie: I assume there must be *something* bad about the hybrid kernel.. otherwise.. why not roll it out to production?
|
| Posted by NHRoel, 09-25-2012, 10:08 AM |
| A lot of business is actually using CloudLinux on production servers. One of the main reason why not everyone use it " It prices money" to use cloudlinux which many people aren't willing to spend.
--Roel.
|
| Posted by mrzippy, 09-25-2012, 10:13 AM |
| We are *already* using cloudlinux on our production servers. I think you missed the point here.
If hybrid 5.x kernel is bug-free and includes all the 6.x features.. then why not make it production kernel instead of "hybrid" kernel? It looks like there is some problem with hybrid kernel, and that makes me think we should only be using production kernel 5.x with our production machines...
Am I incorrect? Is anyone here using hybrid 5.x kernel on their production machines?
|
| Posted by BestServerSupport, 09-25-2012, 10:14 AM |
| Some of the disadvantages of the modular approach are:
1. With more interfaces to pass through, the possibility of increased bugs exists (which implies more security holes).
2. Maintaining modules can be confusing for some administrators when dealing with problems like symbol differences.
|
| Posted by WebHostDog, 09-28-2012, 03:07 PM |
| You can use a kernel which supports cgroups which CloudLinux is using to limit IO. So you basically does not need to upgrade to CentOS 6.
|
| Posted by TravisT-[SSS], 09-28-2012, 03:10 PM |
| Listen to the group that says do a fresh install and then move over and switch over. You will save yourself a lot of problems and headaches.
Install a new hard drive, configure it, move everything over, then boot to it. If done right downtime is as minimal as switching kernels.
|
Add to Favourites 
Print this Article
Also Read
