Synchronize configuration of two XO for high availability
-
What are the best practices regarding achieving a synchronized configuration of multiple Xen Orchestra machines?
(I'm thinking a little bit of two domain controllers that serves one AD configuration, achieving high availability ...)
I have two separate machines serving XO, offering redundancy when updating and rebooting for instance. I'm somewhat concerned about making mistakes when time is of essence and all the ACL configurations is needed to be done twice, every time.
Is it Settings > Export and a job that copies and imports on the other machine, or is there another, "more refined", way?
They need to be able to get users / groups using LDAP.
Lurker for quite some time. First post I guess. Thank you so much for maintaining Xen / XCP-ng. Looking forward to supporting the project moneywise as soon as I start making enough money with my newborn business.
-
Hi,
No need. XO configuration doesn't move a lot. All host objects (pools, VMs, etc.) are only in memory and not stored anywhere else. All in all, XO is relatively stateless, meaning there's no point on getting "HA" between to XOs. Saving the metadata in a backup job every xx hours if you are doing a LOT of changes, otherwise, once a week is probably more than enough.
If you lose your XO, just redeploy a new XOA (or reinstall it) and import the saved config, that's it!
-
@olivierlambert Hi and thank you for your answer.
I don't understand why you say high availability for XO is not a point. I've set up two XOs for the XCP-ng hosts we have. Whenever I want to do service on one of the XO machines, like installing updates or rebooting for other purposes, the management GUI is instead made to be presented from the other XO machine.
We really want this, works great for us. The only question to us here is if the export and import feature is to be used and relied on or if there is another way to synchronize the two - as the changes to the configuration increases.
Redeploying a new machine whenever a primary XO goes down is not good enough strategy for our business model as that could easily take hours or even days if the right people are not present should the XO go down for instance.
-
The essence of XO is to run at one place and manage your entire infrastructure from there. Not to multiply the number of XO instances.
I'm not sure to get which use case require to get HA for XO without getting any pro support behind it?
-
@olivierlambert Well, for now, I'm mostly playing around with this tech to teach myself about it, gaining experience and coming up with ideas. Maybe someday I'll be able to return some profit with the business I'm building / trying to figure out (hosting game servers). Then I'll sign up for some Pro support - and maybe then we can work on the concept of HA XO
I understand what you're saying. I think I can come up with a good enough way of achieving the HA I'm looking for.
Anyway, thanks for takning the time to respond. I understand your schedule is busy.
