Switch unavailable

[Jason Lamping]

I have a WS-C3850-12S that is providing fiber connectivity between my data center core and data closet across my campus.  Twice in the last 2 days that switch has become unresponsive, meaning I am unable to remote in to it, and it will not pass traffic (voice, data, wireless) but I can log in to a switch that sits behind it.  I had to restart the switch 3 times yesterday within the course of 10 minutes and it eventually stayed up for 12 hours, it then became unresponsive.  I was able to telnet to it from the directly connected switch, but there was nothing in the logs.  This morning i upgraded the code from 03.03.03E to .03.07.04E.  The switch has been stable for 5 hours now.  My concern is that it will happen again, and I am unable to find root cause, and I dont have a great next step except replace hardware which will take a few days to procure.  Any thoughts?

[Reza Sharifi]

It is probably still a good idea to open a ticket with TAC, as they may have seen cases like this before and can tell you if this is bug and if the new IOS is free of that bug.  Of courses, if you would have been better if you had opened a case before the upgrade.

HTH

[Jason Lamping]

Yes the switch behind it is on the same MGMT segment.  This switch is not running any routing protocols.

[Philip D'Ath]

I would never run 03.03.03E in a production environment.  Those earlier versions had lots of bugs.

Personally, I would run 03.06.04.E as it is a "gold star" release for most of the 3850 series.  However I'm sure 03.07.04E will be an improvement over 03.03.03E.

You say you were not able to access it remotely, but could access it from a directly connected switch.  Is this device running a routing protocol?  Any chance it simply lost the route back to you for some reason?

[paul driver]

Hello

When you did managed to telnet to it, any statistics on cpu /memory history show up, Was this checked?

If this is acting as a distribution switch and you say a switch sitting behind it was accessible (I am assuming via this distribution switch and not another path and on the same MGT subnet),

Then It does sounds like a possible buggy ios.

res

Paul