[SOLVED] Cant create ceph with fusion-IO ioDrive2

[kantalization]

Dear all,

I used proxmox 5.1 and cluster 2 nodes. Then i want to create ceph storage using /dev/fioa as per i installed the fuison-IO with drivers. Try with pveceph createosd /dev/fioa show unable to get device info for 'fioa'. Previously i have try with ceph-disk zap /dev/fioa and ceph-disk prepare /dev/fioa it seems ok operation was finished, but it take no effect still got unable to get device info for 'fioa'.
Any ideas to fix this?

Many thanks,

 

[ronweister]

are you using the latest drivers?

 

[ronweister]

did you compile the drivers? I had to.

 

[kantalization]

Ya everything is oke except cephosd creating,previously i mounted the fusion io to directory and made it disk images its oke

 

[ronweister]

That is the strangest thing ever. Can you send me the logs for ceph-deploy....I am using fusion iodrives with ceph but on external cluster and they work fine.

 

[kantalization]

i got error from this step : pveceph createosd /dev/fioa, unable to get device info for 'fioa' , i cant moving forward before this step finished

 

[Alwin]

i got error from this step : pveceph createosd /dev/fioa, unable to get device info for 'fioa' , i cant moving forward before this step finished

May you please share the solution with us. Maybe there is something where our pveceph can be improved.

 

[kantalization]

I think its the hardcoded of proxmox disk extention,fioa as fusion-io disk not belonging?now i just try wirh drbd for storage but it takes a long time to sync and i dont know how stable to do cluster node

 

[Alwin]

unable to get device info for 'fioa'.

If this message is from ceph-disk, that it can't handle the fioa device, then it is not likely a issue in the PVE code. As an alternative there is the ceph-volume to create OSDs but its handling is not integrated into our tooling yet. It still should work though.

 

[kantalization]

its mean that pveceph tools work under ceph-disk utility, when i got this error should be try to use ceph-volume?it can create bluestore to share the disk for HA in proxmox?please advice thank you

 

[Alwin]

its mean that pveceph tools work under ceph-disk utility, when i got this error should be try to use ceph-volume?

As an alternative means, yes.

it can create bluestore to share the disk for HA in proxmox?

Ceph is a distributed, shared storage. You need at least three nodes for Ceph and HA.

 

[kantalization]

Thanks alwin for explanation,so it couldn't reach for 2 nodes?when i try with 2 node got no quorum when one node is down,but in each other i user proxmox 3 with 2 noden with fences it running till now can you explain again?thanks

 

[Alwin]

when i try with 2 node got no quorum when one node is down,but in each other i user proxmox 3 with 2 noden with fences it running till now can you explain again?

Come again? I don't quite understand. For HA you need at least three nodes. OSD server could be only two of course.

 

[kantalization]

Hi Alwin,
Previously i had HA proxmox(version 3.1) cluster 2 nodes with fences and iscsci disk mounted on it running well till now. Why on proxmox 5 version must be have at least e nodes?When i tried with 2 nodes clusterr iscsci disk mounted(without fences exactly)and then i tested the HA got error no quorum dead words,can u explain again why and what the different?

 

[kantalization]

*At least 3 nodes

 

[Alwin]

Quorum, for any HA solution.
https://en.wikipedia.org/wiki/Byzantine_fault_tolerance