Understanding how migration and failover work in Proxmox/Ceph cluster?

Discussion in 'Proxmox VE: Installation and configuration' started by victorhooi, Jul 10, 2019.

  1. victorhooi

    victorhooi Member

    Joined:
    Apr 3, 2018
    Messages:
    132
    Likes Received:
    6
    Hi,

    Say I have a Proxmox cluster, with Ceph as the shared storage for VMs. Our VMs are mostly running Windows, and clients access them via RDP.

    To confirm - migrating a running VM from one node to another should be fairly quick, and the VM stays running for the whole period - so a RDP session might not drop, right

    However, if a node suddenly fails - what happens? Eg. what happens to the memory contents of the VM on the failed node? Is there any way to get it to shat it migrates across seamlessly?

    Thanks,
    Victor
     
  2. Alwin

    Alwin Proxmox Staff Member
    Staff Member

    Joined:
    Aug 1, 2017
    Messages:
    2,345
    Likes Received:
    212
    No, if a node failed the VM/CT of that node will be started on another node.
     
    Stop hovering to collapse... Click to collapse... Hover to expand... Click to expand...
  3. victorhooi

    victorhooi Member

    Joined:
    Apr 3, 2018
    Messages:
    132
    Likes Received:
    6
    Right - so it will be a new boot of that VM.

    Curious - is there any method, or scenario under which it could be seamless migrated over, without a restart? Is such a thing possible under Proxmox (or elsewheere)?
     
  4. Alwin

    Alwin Proxmox Staff Member
    Staff Member

    Joined:
    Aug 1, 2017
    Messages:
    2,345
    Likes Received:
    212
    If you mean something like fault-tolerant (hot-standby), then no. Code for this landed in qemu, but to my knowledge it is not production ready yet.
     
    Stop hovering to collapse... Click to collapse... Hover to expand... Click to expand...
  5. fabian

    fabian Proxmox Staff Member
    Staff Member

    Joined:
    Jan 7, 2016
    Messages:
    3,283
    Likes Received:
    506
    for some services there is also the option of doing such hot standby/failover on the application level - e.g. some database solutions offer it. it is very costly though, since you basically have >2x the load all the time. for most scenarios, shared storage for having a readily available cold standby (with automatic failover with our HA stack) is more than enough.
     
    Stop hovering to collapse... Click to collapse... Hover to expand... Click to expand...
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice