• VMs over NFS, who knew?

    From poindexter FORTRAN@1337:3/178 to All on Fri May 22 08:01:11 2026
    On one of my podcasts, the subject was homelab backup. One of the hosts mentioned that he didn't need high-availability features in his 2-node
    Proxmox cluster because he ran his VMs from NFS.

    Wait, you can do that?

    I run my VMs out of local storage on 2 nodes. I'm not running anything production-like, so if I lost a node (which I did recently) then going
    back to a backup or a snapshot onto the remaining node is fine.

    Leaving the VM on NFS means if I lost a node, I could just recreate a
    node on the remaining node and point to the VM disk in NFS. Interesting.

    I have an NFS share on Synology mounted on both nodes (they back up
    there) so I tried migrating the disk volume to it, and it works! A
    little slower than local, but certainly usable.

    The only issue is the noise. My Synology is chattery, and when hitting
    the BBS VM over NFS, it makes a lot of noise.

    I have a plan to migrate my NFS into a storage area, I'm also thinking
    about buying the 10GB ethernet adapter for Synology and connecting it
    and my primary node together over 10GBe or 2.5GBe - whichever cost I can stomach for the switch and a couple of PC adapters.



    ... Do nothing for as long as possible
    --- MultiMail/Win v0.52
    * Origin: realitycheckBBS.org -- information is power. (1337:3/178)
  • From MeaTLoTioN@1337:1/101 to poindexter FORTRAN on Fri May 22 20:35:06 2026
    On 22 May 2026, poindexter FORTRAN said the following...

    On one of my podcasts, the subject was homelab backup. One of the hosts mentioned that he didn't need high-availability features in his 2-node Proxmox cluster because he ran his VMs from NFS.

    Wait, you can do that?

    I run my VMs out of local storage on 2 nodes. I'm not running anything production-like, so if I lost a node (which I did recently) then going back to a backup or a snapshot onto the remaining node is fine.

    Leaving the VM on NFS means if I lost a node, I could just recreate a
    node on the remaining node and point to the VM disk in NFS. Interesting.

    I have an NFS share on Synology mounted on both nodes (they back up
    there) so I tried migrating the disk volume to it, and it works! A
    little slower than local, but certainly usable.

    The only issue is the noise. My Synology is chattery, and when hitting
    the BBS VM over NFS, it makes a lot of noise.

    I have a plan to migrate my NFS into a storage area, I'm also thinking about buying the 10GB ethernet adapter for Synology and connecting it
    and my primary node together over 10GBe or 2.5GBe - whichever cost I can stomach for the switch and a couple of PC adapters.

    Shared storage for hypervisors is a well-established approach - VMware, Proxmox, etc have been doing it for years. It makes VM migration, HA/failover and moving workloads between hosts much easier since the disks are already accessible from both nodes.

    NFS on a Synology is a perfectly reasonable homelab solution. If you're finding it a bit slow or noisy, 10GbE (or even dedicated 2.5GbE for storage traffic) can help a lot. SSD cache / SSD-backed storage can also reduce the chatter.

    Ceph is another good shared storage option if you ever want to go more distributed and remove the single NAS dependency, though it's generally a bit heavier to run/manage than simple NFS.

    ---
    |14Best regards,
    |11Ch|03rist|11ia|15n |11a|03ka |11Me|03aTLoT|11io|15N // @meatlotion:erb.pw |10S|02SBBSS|08-|10M|08-|100|020001 |10C|02ertified |10B|02BS |10S|02YSOP

    |07ÄÄ |08[|10eml|08] |15ml@erb.pw |07ÄÄ |08[|10web|08] |15www.erb.pw |07ÄÄÄ¿ |07ÄÄ |08[|09fsx|08] |1521:1/158 |07ÄÄ |08[|11tqw|08] |151337:1/101 |07ÂÄÄÙ |07ÄÄ |08[|12rtn|08] |1580:774/81 |07ÄÂ |08[|14fdn|08] |152:250/5 |07ÄÄÄÙ
    |07ÄÄ |08[|10ark|08] |1510:104/2 |07ÄÙ

    ... Old musicians never die. They just decompose!

    --- Mystic BBS v1.12 A49 2023/04/30 (Linux/64)
    * Origin: thE qUAntUm wOrmhOlE, rAmsgAtE, uK. bbs.erb.pw (1337:1/101)
  • From paulie420@1337:3/129 to poindexter FORTRAN on Fri May 22 19:11:10 2026
    On one of my podcasts, the subject was homelab backup. One of the hosts mentioned that he didn't need high-availability features in his 2-node Proxmox cluster because he ran his VMs from NFS.

    2-node PVE cluster??? But if one goes down you won't have quorum - you have to add a 3rd for sure! I use a Raspberry Pi, unless I have a proper 3rd PVE node... 3 are needed for quorum, baby!!

    Leaving the VM on NFS means if I lost a node, I could just recreate a
    node on the remaining node and point to the VM disk in NFS. Interesting.

    I also didn't realize I could use NFS shares as VM/LXC drives - super cool, but I don't think as fast as proper HA disks...



    |07p|15AULIE|1142|07o
    |08.........

    --- Mystic BBS v1.12 A49 2024/05/29 (Linux/64)
    * Origin: 2o fOr beeRS bbs>>>20ForBeers.com:1337 (1337:3/129)
  • From MeaTLoTioN@1337:1/101 to paulie420 on Sat May 23 14:31:38 2026
    On 22 May 2026, paulie420 said the following...

    On one of my podcasts, the subject was homelab backup. One of the hos mentioned that he didn't need high-availability features in his 2-nod Proxmox cluster because he ran his VMs from NFS.

    2-node PVE cluster??? But if one goes down you won't have quorum - you have to add a 3rd for sure! I use a Raspberry Pi, unless I have a proper 3rd PVE node... 3 are needed for quorum, baby!!

    Unless you're doing proper HA stuff, quorum prolly isn't _that_ important, but yeah you don't need another "node" to make quorum, just a pi will suffice.

    Leaving the VM on NFS means if I lost a node, I could just recreate a node on the remaining node and point to the VM disk in NFS. Interesti

    I also didn't realize I could use NFS shares as VM/LXC drives - super cool, but I don't think as fast as proper HA disks...

    Yeah tho you prolly wouldn't wanna use it as the only vm disk source, maybe just a second vm disk to back stuff up in the guest perhaps, unless it was fast NFS with SSD backing it, 10Gb/s fast would probably be ok. If I wanted to make it full on, I would definitely have a ceph cluster for the vm disks and just use the hypervisors for the compute.

    ---
    |14Best regards,
    |11Ch|03rist|11ia|15n |11a|03ka |11Me|03aTLoT|11io|15N // @meatlotion:erb.pw |10S|02SBBSS|08-|10M|08-|100|020001 |10C|02ertified |10B|02BS |10S|02YSOP

    |07ÄÄ |08[|10eml|08] |15ml@erb.pw |07ÄÄ |08[|10web|08] |15www.erb.pw |07ÄÄÄ¿ |07ÄÄ |08[|09fsx|08] |1521:1/158 |07ÄÄ |08[|11tqw|08] |151337:1/101 |07ÂÄÄÙ |07ÄÄ |08[|12rtn|08] |1580:774/81 |07ÄÂ |08[|14fdn|08] |152:250/5 |07ÄÄÄÙ
    |07ÄÄ |08[|10ark|08] |1510:104/2 |07ÄÙ

    ... Confucius say: "Man who runs behind car gets exhausted"

    --- Mystic BBS v1.12 A49 2023/04/30 (Linux/64)
    * Origin: thE qUAntUm wOrmhOlE, rAmsgAtE, uK. bbs.erb.pw (1337:1/101)
  • From Mike Powell@1337:3/103 to POINDEXTER FORTRAN on Sat May 23 10:23:17 2026
    On one of my podcasts, the subject was homelab backup. One of the hosts mentioned that he didn't need high-availability features in his 2-node Proxmox cluster because he ran his VMs from NFS.

    Wait, you can do that?

    Yes, if you mean what I think you do.

    I have a Virtualbox instance running an older version of linux. Virtualbox runs on the local machine but the image it fires up lives on my local
    network NFS fileserver.


    * SLMR 2.1a * My computer NEVER locks u ?Ce NO CARRIER
    --- SBBSecho 3.28-Linux
    * Origin: Capitol City Online (1337:3/103)
  • From paulie420@1337:3/129 to MeaTLoTioN on Sat May 23 23:15:10 2026
    2-node PVE cluster??? But if one goes down you won't have quorum - yo have to add a 3rd for sure! I use a Raspberry Pi, unless I have a pro 3rd PVE node... 3 are needed for quorum, baby!!

    Unless you're doing proper HA stuff, quorum prolly isn't _that_
    important, but yeah you don't need another "node" to make quorum, just a pi will suffice.

    :P I disagree - even if we're not super worried about the systems, I always want to know my PVE won't go offline if one of the nodes dies. :P



    |07p|15AULIE|1142|07o
    |08.........

    --- Mystic BBS v1.12 A49 2024/05/29 (Linux/64)
    * Origin: 2o fOr beeRS bbs>>>20ForBeers.com:1337 (1337:3/129)
  • From poindexter FORTRAN@1337:3/178 to MeaTLoTioN on Sun May 24 09:08:08 2026
    MeaTLoTioN wrote to poindexter FORTRAN <=-

    Shared storage for hypervisors is a well-established approach - VMware, Proxmox, etc have been doing it for years. It makes VM migration, HA/failover and moving workloads between hosts much easier since the
    disks are already accessible from both nodes.

    Done it for years with VMWare, knew about ceph, didn't think about NFS.

    NFS on a Synology is a perfectly reasonable homelab solution. If you're finding it a bit slow or noisy, 10GbE (or even dedicated 2.5GbE for storage traffic) can help a lot. SSD cache / SSD-backed storage can
    also reduce the chatter.

    There's a 10GBe adapter for the Synology, it wouldn't take a lot of work
    to create a faster network for the VMs and NAS. What I really need to do
    is wire a shelf in my storage area with ethernet and move the NAS out of
    my office. It'd be cheaper than replacing the drives nowadays and quiet
    my office down a lot.






    Ceph is another good shared storage option if you ever want to go more distributed and remove the single NAS dependency, though it's generally
    a bit heavier to run/manage than simple NFS.

    ---
    |14Best regards,
    |11Ch|03rist|11ia|15n |11a|03ka |11Me|03aTLoT|11io|15N // @meatlotion:erb.pw
    |10S|02SBBSS|08-|10M|08-|100|020001 |10C|02ertified |10B|02BS |10S|02YSOP

    |07ÄÄ |08[|10eml|08] |15ml@erb.pw |07ÄÄ |08[|10web|08] |15www.erb.pw |07ÄÄÄ¿
    |07ÄÄ |08[|09fsx|08] |1521:1/158 |07ÄÄ |08[|11tqw|08] |151337:1/101 |07ÂÄÄÙ
    |07ÄÄ |08[|12rtn|08] |1580:774/81 |07ÄÂ |08[|14fdn|08] |152:250/5 |07ÄÄÄÙ
    |07ÄÄ |08[|10ark|08] |1510:104/2 |07ÄÙ

    ... Old musicians never die. They just decompose!

    --- Mystic BBS v1.12 A49 2023/04/30 (Linux/64)
    * Origin: thE qUAntUm wOrmhOlE, rAmsgAtE, uK. bbs.erb.pw (1337:1/101)

    --- MultiMail/Win v0.52
    * Origin: realitycheckBBS.org -- information is power. (1337:3/178)
  • From poindexter FORTRAN@1337:3/178 to paulie420 on Sun May 24 09:08:08 2026
    paulie420 wrote to poindexter FORTRAN <=-

    2-node PVE cluster??? But if one goes down you won't have quorum - you have to add a 3rd for sure! I use a Raspberry Pi, unless I have a
    proper 3rd PVE node... 3 are needed for quorum, baby!!

    I've thought about taking my Pi4 and making it a 3rd node. Wish it had
    enough horsepower to run PBS.

    The one time I had a node failure, I just restored the BBS from backup
    to my remaining server. Wasn't a huge effort.



    --- MultiMail/Win v0.52
    * Origin: realitycheckBBS.org -- information is power. (1337:3/178)
  • From paulie420@1337:3/129 to poindexter FORTRAN on Sun May 24 16:18:02 2026
    2-node PVE cluster??? But if one goes down you won't have quorum - yo have to add a 3rd for sure! I use a Raspberry Pi, unless I have a proper 3rd PVE node... 3 are needed for quorum, baby!!

    I've thought about taking my Pi4 and making it a 3rd node. Wish it had enough horsepower to run PBS.

    My PVE quorum node is a Pi4 - it happily serves its 'quorum' purpose AND runs a Pi-Hole and Pi-VPN nicely...

    For my PBS, I just use some 200X Dell tower - enough SATA to handle large backup drives and I used its GPU card for something else...



    |07p|15AULIE|1142|07o
    |08.........

    --- Mystic BBS v1.12 A49 2024/05/29 (Linux/64)
    * Origin: 2o fOr beeRS bbs>>>20ForBeers.com:1337 (1337:3/129)