Unsolved

1 Rookie

 • 

1 Message

7

January 16th, 2026 15:17

CSI Driver IPv4 IPv6 Stack supportability - Powerstore - multiple questions

Question1
  1. A date  when IPv6 will be supported for the CSI driver in connection with Powerstore 
Question2
  1.  A) Does Powerstore communicate using the CSI driver via  Management network only or via Management and Storage Network (iSCSI, NAS)?
    B)  In other words  do  Management and Storage Networks need to run in IPv4 or  do we support having only Management in IPv4 and Storage in IPv6? 
    The Feedback from Vyshak ("I believe the CSI driver communicates both via the management and storage network. I don't have answers to questions on IPv6" ) indicates via Management and Storage Network, answering 2)A) 
    Is this the final answer and what about 2) B) ?  Based on 2)A)  I would think then 2)B)  should answer that  Management and Storage Networks need to be on IPv4. 
Question3
      3)  Finally the customer provided 3 Scenarios based on the picture below. And we need a feedback on Dell support for those 3 scenarios:
There are 02 parts to be checked here, 1st is the CSI communication (Management traffic) and 2nd is the iSCSI data traffic, as depicted below:
Further see picture here with additional comments: 
  1. Worker nodes (0-3) hosting CNF have 02 physical ports on each server for iSCSI multipath A & B, and this communication will be happening via a dedicated IPv6 storage pools for path A & B, and this storage pool spans across the 04 worker nodes. (dashed redline) 
  2. CSI communication will be happening via the controller nodes of the USM cluster via the default OAM network which is also on IPv6 (solid red line)
  3. USM cluster is an IPv6 only cluster i.e., K8S is only operating on IPv6 stack.
Based on the recent communication, it has been mentioned that Dell CSI doesn't support the CSI communication (Point 2) via IPv6. This means the networking on caas layer is also impacted. Here are the options/scenario's to be checked: 
#Scenario A:
Convert USM subcloud to Dual stack (i.e.., K8S is operating on both IPv6 and IPv4 and there are dedicated pod, service and cluster subnets for both IPv4 and IPv6) , so the CSI can communicate with IPv4 endpoint of K8S API server. Is this scenario supported i.e., K8S is dual-stack and CSI is IPv4 only ? And when CSI is over IPv4 network, can the iSCSI data path still be IPv6 ? 
  
(My comment:  IPv4 for Powerstore Management + IPv6 for Powerstore Storage, here iSCSI )
#Scenario B:
Same as Scenario A, but in this case both CSI and iSCSI are over IPv4 networks.
NOTE: In both scenarios A & B the caas layer is Dual stack.
(My comment:  IPv4 for Powerstore Management + IPv4 for Powerstore Storage, here iSCSI )
#Scenario C:
On IPv6 only subcloud of WRCP, the K8S API server is hosted on IPv6 and all pods, services only get IPv6 addresses. WRCP provides the option to provision additional IPv4 pools also on IPv6 subcloud. So, when we deploy the Dell CSI on IPv6 only subcloud, the CSI pods get IPv6 addresses, and the PowerStore array configured as IPv4 endpoint can be reachable via the IPv4 egress interface on the controllers. So in summary, the CSI pods and K8S is on IPv6, but the Storage array is on IPv4 network. Is this scenario supported ? 
(My comment:  not 100% clear as only mentioned the storage array on IPv4 not differenciating between Mgmt/iSCSI  - so  this  seems IPv4 for Powerstore Management +  IPv4 for Powerstore Storage, here iSCSI  but I could have got this wrong)
No Responses!

Top