SOLVED Question about SuperMicro support

[spiceygas]

I'm looking to build a FreeNas file server. Considering a system with the SuperMicro X10DRi-T4+ motherboard. When looking at their official OS support page, that board supports FreeBSD 10.0.

It appears that the current version of FreeNas uses FreeBSD 11.3. Should I be concerned about compatibility? Have others had problems using SuperMicro boards on newer versions of FreeBSD than are listed on the official compatibility page?

(Admittedly, the board is a few years old so they probably just didn't go back to test it with a new version...?)

 

[wl714]

The board has C612 chipset and Intel X540 NIC, all supported devices, I won't expect any compatibility issue. However, if the reason for picking this board is to link aggregate the 4 ports (with a managed switch) to expect near 40GbE networking performance, you might be disappointed in most cases. The statement came from reading other posts in the forum, not personal experience. But enough for me to go with QSFP+ directly dropping the original plan of linking multiple 10GbE. If you plan link aggregation, lacp better to do more forum reading to know what to expect.

 

[Yorick]

iSCSI multipath would be the answer, instead of LACP LAG.

To the original question: X10 and X9 generation Supermicro boards work fine with FreeBSD 11.3 and for that matter 12.1. I don’t expect that’ll change with 13, either. These are very well liked as hobby servers.

 

[spiceygas]

Both very helpful responses. Thank you.

The board has C612 chipset and Intel X540 NIC, all supported devices, I won't expect any compatibility issue.

To the original question: X10 and X9 generation Supermicro boards work fine with FreeBSD 11.3 and for that matter 12.1. I don’t expect that’ll change with 13, either. These are very well liked as hobby servers.

This is exactly the answer I was looking for. Thank you.

However, if the reason for picking this board is to link aggregate the 4 ports (with a managed switch) to expect near 40GbE networking performance, you might be disappointed in most cases.

iSCSI multipath would be the answer, instead of LACP LAG.

This is for home use, so 1x 10 gigabit will be plenty. I appreciate the warning and management of expectations :)

 

[spiceygas]

However, if the reason for picking this board is to link aggregate the 4 ports (with a managed switch) to expect near 40GbE networking performance, you might be disappointed in most cases. The statement came from reading other posts in the forum, not personal experience.

Incidentally, some googling found an official Intel guide on setting up LACP for this adapter. I'm a home user, so I don't really need it. But others looking at this thread in the future may want to experiment with it. Their guide covers Windows, Red Hat Linux, and ESXi. Not FreeBSD.

 

[Yorick]

The thing about LACP LAG is that it is load sharing, not load balancing. The switch creates a hash of source/destination tuples, and chooses a link accordingly. Some switches can do this at Layer 4, based on not just IP but also port, which can give better results when there are multiple ports in play.

For your bog-standard IP-based hash and multiple clients hitting a machine, received wisdom is that you need a good dozen-ish clients to really get good use of a LAG. With two clients, both might end up on the same link, that depends entirely on their IPs. Sure you can fiddle with IP addressing in your network until you observe clients hitting separate links, but good Scott.