Procurve 2910 weird behavior with DHCP relay

[BDW]

Hello everybody,

we're facing some strange issues on our 2910 switches since end of April.

All the switches behaved correctly on several remote sites.  All sites received a correct IP helper address and everything worked fine for several years.

A few weeks ago we had the first outage that we didn't receive new IP's anymore on 2 sites.  We thought.. ai, issues on DHCP server but other sites received IP's correctly.  After restarting switches, we tried to remove IP helper, reboot switch, add IP helper again, reboot and everything worked correctly again... weird... but OK, it recovered.

Since that moment we see on several sites similar issues.  Most strange thing is that the issue occurs on 3 sites and that it never starts alone but always on 2 sites at once (at least).  We have the same config on 5 sites and 2 other sites didn't have any issue.  All the sites don't have a real relation between each other.  We can always recover by replacing ip helper.

We've been searching and digging but we can't think of a solution or where to look.  Did someone see similar behavior before ? 

Thanks in advance !

[TerjeAFK]

What kind of connection do you have between the sites? Did you try Wireshark on the DHCP server to see if the DHCP request reached the server?

[Vince-Whirlwind]

A bit hard to know where to start, but my first thought would be to rule out STP blocking was lasting longer than the timeout on the DHCP request.

If the DHCP server saw the request, it should reserve a lease against the number in its Offer. Can you see that the Offer has been sent?