The last few generations of HP ProLiant systems have featured an option to share the ILO functionality with the first network interface. I'm in a situation where per-port costs on the switch I'm using the a co-lo facility are fairly high. I can save a port per server by not using the dedicated ILO interface. Is there any disadvantage to doing so? I've run with it on a few systems, and it works, but I'm sure there's a reason it's not the standard. I'd appreciate any insight.
I believe that you need to weigh up the pro's and cons of the sharing the hardware.
Sharing the NIC would allow you to save a little bit of money on your colo costs which is a good thing, however by going down a shared ILO route you should consider:
- Utilisation of the first NIC, if you have a heavy network load how will you access the server via the ILO? If you have a separate NIC then you have the full bandwidth to use to access the server remotely.
- A hardware failure of the first NIC will also nix your ability to remotely manage the server, this may not be something that matters much to you but in my experience (With Dell iDRAC cards) having the second NIC is a benefit when things go really wrong.
At the end of the day I can't give concrete advice, its just a decision you need to make based on your business needs and available funds.
Another disadvantage (don't know if this applies to all iLO versions and all Proliant models) is that when iLO is on shared port, it cannot communicate with host computer. So server can't even ping its own iLO IP address, although all other devices can - this is because of how this NIC sharing works.
I've seen no differences between the two styles expect if you start to max out BW on the NIC, If it saves you co-lo costs then I'd get it except if you think you will run above 90% utilization but at that point you need a better NIC. Although you still need a separate IP to assign to it.
There is another issue as well, in the current iLO 4 (version 2.40 Dec 02 2015) the "Shared Network Port currently does not support IPv6".
The only real disadvantages I can see are that you have your 'OOB' traffic on the same physical wires as your 'In Band' traffic, which could be a security issue - and that one type of traffic could flood and interfere with the other.
Otherwise it sounds workable and in your case preferable.
I've seen many environments with separate switches for ILO/iDrac and other management (like cisco consoles).