Re: Question on iBFT for iSCSI

Re: Question on iBFT for iSCSI

[Subramanian, Sriram (EG Servers Platform SW)]

Any feedback anyone? Is it reasonable to make the change to publish the configured/DHCP provided target IP address rather than the redirected IP to iBFT?

Thanks,
Sriram.

-----Original Message-----
From: Subramanian, Sriram (EG Servers Platform SW) 
Sent: Tuesday, August 23, 2016 7:01 PM
To: 'Ye Ting' <ting.ye@intel.com>; 'Fu Siyuan' <siyuan.fu@intel.com>; 'Wu Jiaxin' <jiaxin.wu@intel.com>
Cc: 'Zimmer, Vincent' <vincent.zimmer@intel.com>; Li, Ruth <ruth.li@intel.com>; edk2-devel@lists.01.org
Subject: Question on iBFT for iSCSI

All,

The IScsiDxe implementation in EDK2 publishes the iBFT with the target IP address to which it connected to, and not the configured IP. These IPs will be the same if there is no iSCSI redirection. However, if the configured target redirects the initiator to another target (as part of iSCSI Login response), the Session data is updated to the new target IP.

To my knowledge, there doesn't look to be any standard/specification governing this, so I believe this is open to debate. 

The reason for this question is that the configured IP address may be that of a server that's doing the redirect to achieve load balancing or high availability in the cluster.

Now if we set the Target IP address in the iBFT to the final IP to which we connected, we lose the redirection capability in case the target at the final IP goes down, and the OS wants to re-establish the connection. 

If we published the original configured IP, then the OS will attempt to reconnect to the configured IP in the iBFT, and will get redirected to another target which is up.

So do you think it makes sense to publish the original configured static IP (or the original IP provided via DHCP) instead of the redirected IP? What are your thoughts on this?

Thanks,
Sriram.

Question on iBFT for iSCSI

[Subramanian, Sriram (EG Servers Platform SW)]

All,

The IScsiDxe implementation in EDK2 publishes the iBFT with the target IP address to which it connected to, and not the configured IP. These IPs will be the same if there is no iSCSI redirection. However, if the configured target redirects the initiator to another target (as part of iSCSI Login response), the Session data is updated to the new target IP.

To my knowledge, there doesn't look to be any standard/specification governing this, so I believe this is open to debate. 

The reason for this question is that the configured IP address may be that of a server that's doing the redirect to achieve load balancing or high availability in the cluster.

Now if we set the Target IP address in the iBFT to the final IP to which we connected, we lose the redirection capability in case the target at the final IP goes down, and the OS wants to re-establish the connection. 

If we published the original configured IP, then the OS will attempt to reconnect to the configured IP in the iBFT, and will get redirected to another target which is up.

So do you think it makes sense to publish the original configured static IP (or the original IP provided via DHCP) instead of the redirected IP? What are your thoughts on this?

Thanks,
Sriram.