proxmox cluster, ceph - dedicated network disconnected

[spark 5]

hi guys, great to be here ... i' m new with proxmox and it looks really great.
i setup an test cluster for one of our customers.
4 server as cluster with ceph as storage.
i use 3 networks, 1 ceph, 1 proxmox ha corosync, 1 for virtual machines.

everything works great.

now i thought, what is happening, if one proxmox node is loosing the connection to ceph network.
nic or switch or cable damaged.

if i would run proxmox corosync through the ceph network, the cluster detect this and move my vm to another node.
but with the dedicated network, ceph tells me, one node is gone, but proxmox does not recognize that.
so i have a vm running, wihout harddisk.

so, for sure i need my ceph storage redundant, but did i miss something?
is there a way to check ceph as cluster resource via proxmox and migrate vm, if local ceph goes down?

or do i' m hunting a theoretical problem?

thanks for your help and for that great forum,
ronny

 

[Alwin]

so, for sure i need my ceph storage redundant, but did i miss something?

Redundant networking is needed in that case.

is there a way to check ceph as cluster resource via proxmox and migrate vm, if local ceph goes down?

Through external monitoring the cluster and storage.

 

[spark 5]

hi alwin and thank you

i tried this via disabling the network link and my testvm stuck.
i also cant migrate it to anoter host:
2018-09-27 08:56:02 migration status: active (transferred 211743119, remaining 4227072), total 554508288)
2018-09-27 08:56:02 migration xbzrle cachesize: 67108864 transferred 0 pages 0 cachemiss 8 overflow 0

so it looks like a critical part ...

what would you say, is the best way for redundant network for proxmox/ceph? any experience?

thanks,
ronny

 

[Alwin]

what would you say, is the best way for redundant network for proxmox/ceph? any experience?

IMHO, a classic active-backup network failover is a good go. But others may disagree.