XO in HA mode



  • Hi - is it possible to run 2 instances of XO managing the same xcp-ng cluster without duplicating the backup jobs etc?


  • Admin

    Hi,

    What's the use case in the end? XOA (and logically XO VMs) are meant to be disposable, because it only stores a very minimal configuration. You can save this config and reimport it quickly if you have a problem with the first one.



  • The issue I faced when the entire pool was destroyed including the xo vm. I had to manually figure out which one of the backup (on nfs share) was xo then import it using xcp-ng, then restore other VM's. So now I have one instance running as VM and other on Intel NUC...


  • Admin

    It's not the easiest way to do that. As I said, because XO doesn't store a lot of info, consider it as disposable. Just deploy/reinstall a new one, and even without your previous settings, you can access the backup. If you have a job with XO metadata backup, then it's even easier.



  • I assumed that, yes I do have metadata, pool, daily delta, weekly full on schedule.



  • Just copy the XOA VM to backup host on every major change in XOA config, in case of event You can start the xoa from backup host, also it's advised to backup the xoa settings to file.



  • I actually want a way to HA this as well. There is a use case for this.

    I use XO for remote access to the Servers and VM's. The issue is, that if the Server with XO or the XO VM goes down, I have no access to any of the other servers or pools.

    At the moment I personally am not too far from work and my home network is a part of my work network, so not too much of an issue to get back in another way. But in January I am heading to Fiji for 2 weeks and will not be able to do anything at all if the XO VM goes down.



  • @eangulus what I am doing is running 2 instances, one on VM, another one in Intel NUC, only connected to one at a time and only one is running the backup, if one goes down, I can use the other by connecting to the pool.


  • Admin

    Then use a fallback XO VM elsewhere. As I said, it's almost stateless, all Xen objects are in RAM.



  • @olivierlambert I don't' think that is a complete solution.

    Yes, you can spin one up whenever needed. But to do that you need access to the servers. So if XO is your only means of access at the time, then how does one just spin up another XOA?

    I don't think it's about having to set up another XOA or about it being easy or difficult to set one up. It is more about the times when it is your only access to the VM's and XCP-ng Servers then what? If you have at least 2 XOA's running then at least if one fails, you still have full access to the servers.

    I like the idea from @shwetkprabhat but in my case I may just run a second XOA on our second Pool and just not run any backups from it.


  • Admin

    That was exactly my point: a fallback XOA running elsewhere when you need, that just sits doing nothing except "in case". No need to have HA for that.



  • @eangulus

    You can always create a CLUSTER with shares storage. Problem solved. It is even possible to create cluster using two nodes - using HA-Lizard.


Log in to reply
 

XCP-ng Pro Support