User guide

DataDirect Networks SFA™ OS 2.0.0 Release Notes Revision A1 | 9
Previously, if you issued a subsystem shutdown on the SFA controller, the VMs
were shut down without warning. Now, a check is made and if VMs are running,
the shutdown command will fail with an error message. This will prevent virtual
machines from being shut down by accident.
o Presenting a VD to all VMs/all Hosts no Longer Fails if Some Presentations Exist
Previously, if you had one or more presentations already set up to one or more
hosts, the following command would fail with an error stating that some
presentations were already defined:
app create presentation host ALL vd 0
This has been resolved. Issuing the above command will now create the
presentation for a given VD to all hosts even if some presentations already exist.
o Ability to See When an Image is Ready
The CLUI app show image command now informs you if the image is ready
for a stack (VM) and can be started.
Previously, it was difficult to know if an image was ready to be used by a
stack/VM. You would have to start the stack to see if it produced an error
message to know if the image was fully allocated or not.
Now this information is provided as part of the output of the app show image
command—including the percent complete.
• The Mellanox firmware on the SFA12K Infiniband ICL is upgraded from 2.10.700 to
2.11.500 keeping it up to date with the front-end Mellanox HCAs on the SFA12K.
4.0 Resolved Issues
These issues have been resolved.
• Errors on one drive cause other drives to go missing
An issue was found in the drive correction firmware where the wrong drive is power
cycled in an attempt at recovery. This would occur when a drive was in an error state
and not responding or when it was physically pulled from the enclosure. This caused
subsequent incorrect power cycles to occur because the drive that was incorrectly
power cycled is now not reporting. As a result, pools will fluctuate between normal,
degraded, non-redundant, and AWL as drives come back from the power cycling.
• Pool inits on systems where pools were created with GUI or API take
excessive amount of time to complete
If the GUI or API was used to create the pools, two commands were sent to the
controller instead of one. This resulted in the DIF init being a non-destructive init
(read-modify-write operation instead of zeroing out the DIF field).
• InfiniBand ICL issues can cause controller watchdog crash
When the ICL HCA experienced issues, a timer in the SFA OS code would go off and
would attempt to correct the HCA. However, this error correction on the InfiniBand
ICL was done by resetting the controller to reset the HCA. This has now been
changed so that the ICL channel will be taken down and back up in an attempt to
correct the issue. If this fails, that channel will be taken down and the ICL will run in
a degraded fashion on the other channels.
• Controller crash after shutdown subsystem abrupt command was issued