User guide

DataDirect Networks SFA™ OS 2.0.0 Release Notes Revision A1 | 12
o If the ULA field has the value "0000000000000000", continue to wait and
not issue any CLI commands besides "show controller".
Here is an example output of the condition where you would want to wait:
$ show controller
*************************
* Controller(s) *
*************************
| Up Time | |Encl| |
…
Idx|Name |Mastership|Locality| D: H: M: S|RP| ID |Idx | ULA |
…
-----------------------------------------------------------------------------------------------
…
0 A PRIMARY LOCAL 0000:20:24:34 1 0001ff0900180000 0 00000001ff0800ac
…
1 B SECONDARY REMOTE 0000:00:00:05 1 0001ff09002d0000 0 0000000000000000
…
Total Controllers: 2
o When the ULA number has something other than all zeros for the remote
controller, it is now safe to issue CLI commands and avoid the
communication problems.
5.2 SFA12K-20E
• VMs can be shown as running when controller hosting those VMs goes down
In certain situations, if one controller in a couplet goes down, the remaining
controller may show the VMs in the offline controller as running. This is because the
remaining online controller cannot update the status of the running VMs on the
offline controller. In this scenario, the online controller continues to display a healthy
running VM status. Once the offline controller comes back online, the correct state of
the VMs will be visible. [DE4409]
• SCSI DID_NO_CONNECT errors
When running GRIDScaler 1.6.1 (OFED 1.5.3.4.0.35), the following erorrs could be
observed.
Apr 12 16:26:42 icndcm10 kernel: sd 8:0:0:1: [sdag] Result:
hostbyte=DID_NO_CONNECT driverbyte=DRIVER_OK
This problem is fixed in GRIDScaler 2.0.0. There is also a field upgrade patch
available. Contact DDN Support if you need this patch.
• Clocksource tsc unstable errors [DE4620]
On SFA12K-20E controllers, error messages similar to the following may be reported
on the running VM with dmesg or in system logging:
Clocksource tsc unstable (delta = -519494463 ns). Enable
clocksource failover by adding clocksource_failover kernel
parameter.
A workaround has been provided for GRIDScaler 2.0.0, so this error should no longer
appear. For other releases, the following workaround can be applied to have the
kernel fallback to HPET mode if the guest OS detects a problem with TSC. In the
kernel parameter list of the VMs, add the following line to the VM kernel boot line in
/boot/grub/grub.conf.
clocksource_failover=1