SOLVED 9.2.1.9 -> 9.3 Lost Intel 1000PT Adapter

[bmh.01]

Have an additional Intel 1000PT Dual Server Adapter in my FreeNAS server for dedicated iSCSI interfaces for my esxi hosts. Was all working perfectly on 9.2.1.9.

Upgraded to 9.3 and whilst FreeNAS can see the adapter just fine (em1/em2) and assign an IP to bother ports, they won't pass any traffic. Can't ping in out and whilst they get a physical link tcpdump doesn't show any traffic being passed.

All i've done is upgrade from 9.2.1.9 nothing has changed configuration wise, i've tried everything I can think of including deleting the adapters from the network config and adding them again, trying them on different subnets etc.

To get the datastores back online i've had to add the two subnets as a vlan of the onboard em0 but obviously this isn't ideal as it's now sharing the same physical interface as everything else.

Anyone have any ideas or had this before? I'm running the CTL iSCSI target if that makes any difference and have been since it first appeared (9.2.1.6/8?)

 

[dlavigne]

Anything related in dmesg (that's your boot hardware probe) or /var/log/messages?

 

[bmh.01]

Nothing helpful. The card gets detected and interfaces assigned (em1 & em2) and I can see the physical link state in ifconfig can pull the cable and it changes. It "works" apart from passing traffic.

Theres an onboard 82579LM controller which gets em0 which works perfectly so it's not like the em driver is missing or doesn't work.

 

[dlavigne]

Any update on this? Anything change elsewhere in the network? Have you tried a packet sniffer to see if packets are being sent out?

 

[bmh.01]

Sorry for not reporting back quicker haven't had chance to look at it again in the flesh before today. Nothing to do with freenas or the hardware it's running on, turns out to be the switch that it is connected to for iscsi traffic the two ports that the adapter was connected to for some reason don't want to pass traffic despite linking up. Added the vlans to the two next avaliable ports and moved the cables over on the switch and voila back as it was before, haven't had chance to reboot the switch yet but I have no doubt that would "fix" the original ports as well.

The switch is a HP 1810-24g v1 which granted does has 196 days uptime on it but still wierd how it co-incided with the upgrade. I never was a big fan of the switch despite the community opinion so I think it's just signed it's own death warrant to replace it with something else. I find it too limited configuration wise anyway.