4 ms·
Intel NICs have had similar problems. One of my circa 2010 servers had the NIC lock up in such a way that passthru traffic to the IPMI was blocked preventing r
by bcrl 3y ago
Intel NICs have had similar problems. One of my circa 2010 servers had the NIC lock up in such a way that passthru traffic to the IPMI was blocked preventing remote management access. I drove 3.5 hours to the data center to power cycle the system. Investigation eventually turned up an errata for exactly that issue. Suffice it to say that I decommissioned that server as soon as I could.
- T3OU-736 3y ago> Intel NICs have had similar problems. One of my circa 2010 servers had the NIC lock up in such a way that passthru traffic to the IPMI was blocked preventing remote management access. I drove 3.5 hours to the data center to power cycle the system. Investigation eventually turned up an errata for exactly that issue. Suffice it to say that I decommissioned that server as soon as I could. In the past life of being a SysAdmin, we strove for independent OOB NICs whenever possible specifically for this sort of edge case. And then used switched type PDUs (instead of just metered) so that remote power-cycling was possible for the eventuality that the IPMI controller had an OOM and no incantation of `ipmitool` would suffice (<caugh> iDRAC <caugh>). Sounds like you weren't able to get all those nice things. Unfortunate. Been there.
- Fischgericht 3y agoIn-Band passthru is a nightmare. Even if your NIC is good, it can cause all sorts of problems. It's okayish if it's done via an Ethernet bridge, but that seems to be the exception. On the ASRock Rack boards I have recently sourced you suddenly get a new PCIe root bridge, and (10G) NIC throughput goes down the drain.