NVME/TCP with vmware vsphere non yet stable non ready for production

[zouhairnador]

please can any one confirme me the stability of nvme/tcp with vmware datastore ? vsphere7 8 or 9 ?
until now all my benchmarks still going bad with no hope that will be ready for production with vsphere.
i still have the VAAI : full clone not supported, and slow performance and unstabilities , and this is with tentacle 20.x .
please help if any one have this in production .
thanks you

[caroav]

It is very stable with ceph tentacle release, and esx 7u3 and beyond. I'm not sure which issues you are having.

[zouhairnador]

is it in production please ?
here first issue :
[root@ESX01:~] esxcli storage core device vaai status get -d eui.999592ebcdba4429ad87e6b500e4fbfd

eui.999592ebcdba4429ad87e6b500e4fbfd

VAAI Plugin Name:

ATS Status: supported

Clone Status: unsupported

Zero Status: supported

Delete Status: supported

[root@ESX01:~]

meaning you cannot offload the VM clone operation to the VAAI storage of ceph , the operation takes houres because the esxi taking the load , deferent than all the storage array technologies .
the nvmetcp is crashing each time (too much load) .

if you dont have hose issues please send me the yaml ile of the depolyement and configuration of your working ceph cluster .
thank you in advance

[caroav]

The fast copy is being implemented now, and will be available soon. The way that copy works now should not be a big difference in terms of overall performance because in any case the data read/write in ceph is distributed across the nodes/network. Please provide the following details:

• ceph and gw version
• how many gws are you deploying in the group?
• in case you say that gw is crashing, please provide gw logs

[zouhairnador]

1. the full clone i'm talking about its from esxi vsphere side not the internal rbd clone, the esxi not offloading its operations of clonning to the storage like all other storage technologies, so the esxi not offloading to ceph , esxi 7 he the clone themself so it take houres to complete because of this unsupported function (only on nvme/tcp) :

---

[root@ESX01:~] esxcli storage core device vaai status get -d eui.999592ebcdba4429ad87e6b500e4fbfd
eui.999592ebcdba4429ad87e6b500e4fbfd
Clone Status: unsupported

but in iscsi its showing supported , and working ,
so give me the version that correceted this bahavioure on nvme/tcp please .

2. Its ceph 20.x tentacle .
3. the default groupe with 3 iscsi gw inside 3 ceph nodes (all roles on the 3 ndoes , mgr, mon , nvme/tcp....)
   4)i will provide with logs yes

[zouhairnador]

are my previous information usables for you to answer to me please?
or you need more info?
have you any production environement exemple like mine? (vspehre 7/8 with ceph tentacle 20? ) is its working good?