UCS-FI-6296UP Webpage Freezing

[allen.george]

Hi All,

We have a pair of 6296UPs at one of our sites and the website on the device freezes almost every time. When you try to log into the device, you get to the page to select UCS Manager or the KVM. Clicking UCSM often doesn't load and sits on the loading screen infinitely. Pressing F5 or refreshing the page will get you in after some period of retrying (usually 10 or 15 times) to the login page. When you log in with credentials it does the same thing. Sometimes it hangs and the browser asks to close the page, sometimes it lets you in (very rarely) and other times it tells you that you have a session open already. After a few hours of trying to get in you can usually get in but it shouldn't be like this obviously.

This appliance is running 4.1.2b. I cannot open a ticket because the hardware bundled with this is no longer under support because the hardware is end of life so we did not try to renew support on the FIs separately.

Is there a webserver service that should be restarted running on the FIs? We have production workload running on this hardware so whatever we do needs to be non-disruptive to the C240s that are attached.

[Steven Tardy]

The few times I've seen something similar to this were "something upstream". Either a firewall or FI management port cabling or upstream MTU.

Some recent-ish version of UCSM (maybe 4.0 ?) added allowing to restart only the http process from the CLI (SSH) with commands like:

connect local
pmon stop httpd
pmon start httpd

This won't fix the issue if the issue is NOT the UCS FI web service.

Troubleshooting this usually requires simultaneous packet captures (client side and UCS FI side) to see if packets are being retransmitted or what is going on (packet 27 sent by FI but never seen by client. . . . not a UCS issue, usually).

[allen.george]

Thanks Steven! I'll try the restart because this started after an upgrade from 3.1.2 to 4.1.2b.

[allen.george]

Unfortunately, that didn't seem to fix it. This problem started after we did our upgrade last year and was ignored until now. /shrug I guess it is what it is.

[Steven Tardy]

Is this both Fabric Interconnects or just one of the two?
(The subordinate should show the first page and attempt to login but then say "UCSM GUI is not available on secondary node".)

Does the first page work or does it only fail after logging in and attempting to load the full GUI?

Can also change which FI is primary (has the Virtual/Cluster IP) to see if the "other" FI works:
https://community.cisco.com/t5/unified-computing-system-discussions/ucs-fabric-interconnect-after-failover-primary-subordinate/td-p/2281245

Double check:
1) the cables of the FI management port and
A poorly seated cable could negotiate at a non-optimal rate.
2) the configuration on the uplink switch.
If the switch hard-codes the speed/duplex that could cause issues.

[allen.george]

I have the networks team checking those items on the uplink ports. The login page works on both of them, which seems weird. I'm going to try a failover with a change ticket open next week.