Good afternoon, I understood from the information that DCM supports Live Migration between different data centers? if so, how does it work? could I migrate databases without stopping writing to them?
Live Migration
- Thread starter HacTpoeHuE
- Start date
Yes this is possible.... but when the VM is finaly switching from one DC to the other, there is a brief downtime.... something between 200 and 5000ms.....
even with different file stores? is there an article where I can read how it works? I need to transfer a software package consisting of 20 VMs and preferably without stopping
Yes, it is a Live-Migration. It takes its time if you migrate between stores, but it is synched and there is no "delta" when the final switch occurs.
I'd suggest to start with the Migration section in our admin guide, there are more general information as well as a "How it works" paragraph
I don't really understand how, for example, a database will migrate if it is constantly being read and written
As @cheiss wrote, this is descriped in the docs{1}.I don't really understand how, for example, a database will migrate if it is constantly being read and written
1: https://pve.proxmox.com/pve-docs/pve-admin-guide.html#qm_migration
Yes, I read this and it doesn't say much about disks. do I understand correctly that since the migration, proxmox stops working with the disk and saves all changes to ram? there is a lot of old information where it says that this cannot be done without a shared repository. I think I'm confused.
Thank you, it's useful. but it is not completely clear what will happen to the disk on different repositories. I have tested this feature and it works. but I want to understand exactly))I'd suggest to start with the Migration section in our admin guide, there are more general information as well as a "How it works" paragraph
It basically starts by transferring the disk contents and starting a Live disk mirror.
After that, RAM will be transferred - at some point, it will "converge", stopping the source VM for a very brief period (normally way below 1000ms), transferring the rest of the (dirtied) RAM and resuming the VM on the target.
All of this is transparent to the guest itself.
But not in our documentation? It is recommended in clusters, of course, but not required in any way.
You should not try to see problems, where are no problems. PVE-Migration is Datacenter and Enterprise-proven for years now. We do this daily multiples times with SQL-Servers, Mail-Servers and dozens of other systems. There was never any "problem". Stable Network and Storage is a requirement...