UCS blade servers are rebooting automatically when i disconnect the connectivity to MDS & core switches from FI

[Gonuguntla Gopi]

Dear All,

I have an infrastructure running with an standalone FI & with 8 blades.

I have configured service profiles & every thing is working fine.

Issue:

When i remove the connectivity on the MDS, network core switch, MGMNT Switch from FI, the servers are getting rebooted automatically....Feeling surprise.

I have re-verified the configuration....i didnt see any loop holes.

Experts please provide me the solution.

Looking forward to hear from you.

Regards,

Gopi G

[Keny Perez]

Do your servers boot from local disks or remote storage?

-Kenny

[Gonuguntla Gopi]

Keny,

Servers boot from local storage.

Regards,

Gopi G

[Keny Perez]

So the servers boot from the local disks and you have a KVM window open to monitor the servers and you take down all uplink ports (ethernet and FC) and the servers you have a KVM open for are rebooting?  Or what behavior/error message makes you think that they are rebooting when the uplinks are down?

-Kenny

[Gonuguntla Gopi]

Even i am surprised with this behavior.

 I have just unplugged the FC uplinks  from FI-MDS Switches& servers got rebooted. So i thought its an issue with the configuration.

But one day I unplugged the Ethernet uplinks from FI-Core switch & servers got rebooted.

Again once the servers came up & i unplugged the MGMNT Port....Then also servers got rebooted.

I have checked the configuration & found no errors.

---Gopi G

[Walter Dey]

What OS ? Blue screen / purple screen ?

[Gonuguntla Gopi]

Walter,

Its Microsoft Hyper -V, Machines are not getting hanged.

the problem is only when i unplug the connectivity of LAN/SAN to FI..

---Gopi G

[Walter Dey]

Do you have blades, without a OS installed (or service profile associated) ? and do they boot as well

Do you have only Hyper-V or other OS as well

What kind of network luns (IP, FC based) has your Hyper-v nodes ? I assume it's not only local disk based ?

[Gonuguntla Gopi]

I have installed hyper v & virtual machines are running on the blades already. the LUNS are FC based. There were zero errors in the configuration of the Service profile & all the blades are running well & good. The blades are installed with hyper V on local disks of each blade.

Yes, your assumption is correct.

Regards,

Gopi G

[Walter Dey]

Hi Gopi

Q. why are you not testing the HA Feature of UCS, having a dual fabric, with Failover if one would fail; your Scenario with both fabrics failing might be not so realistic.

Regarding ....LUNS are FC based..... if you kill any communication on both fabrics, loosing the paths to the luns, might make your OS inoperable, and finally Crash ?

I can hardly believe that this is a intrinsic UCS Hardware blade issue ?

Walter.

[Gonuguntla Gopi]

We tried and tested it  & there was no error message....

[Keny Perez]

but my question is if you are looking at KVM or how are you seeing the server get rebooted?

What server model and what UCSM version?

-Kenny