I was thinking of setting up something similar. Would it work with ESXi?
The SAS3041E-R "works" without any hacks or tweaks in ESXi -- That is to say, it works out-of-the-box as of 3.5.0 (build-123629). I have the PCI Express version of this controller attached to two Hitachi 1TB (3.0gb/s) drives.
I have noticed the following caveats:
When I install ESXi onto a RAID-1 (mirror) configuration, where ESXi resides on the first several hundred megabytes of the disk, it seems that every three or four reboots, the array is degraded and/or out-of-sync.
The system health indicators in VirtualCenter detect the alarm condition on the card, warning me that the array is degraded, and which disk is out of sync -- so it's at least talking to the controller and picking up on that.
Syncing the array, as far as I can tell, needs to be done in the controller's BIOS menu -- takes about 8 or 9 hours to sync the two disks offline. I have read other reports that you must also go into the controller's BIOS menu to replace a failed mirror disk, and that it takes about as long.
There are drivers and software utilities for Windows and Linux, which will monitor / repair the arrays, but these are obviously not included in stock ESXi, and I haven't dug further (yet) to see whether they can be installed in the service console under ESX....
(Yes it was a new install with no VMs. I had nothing invested in its data, but I did want to know how long this would take in the "real world" -- It's a real pain in the a** when it does this randomly, and frequently for no apparent reason).
I updated my system BIOS (Dell Poweredge 830, now has rev A04) and the firmware on the LSI card. I have made the following observations:
1) If I install ESXi on USB or the PERC (SCSI), dedicating my entire SATA raid-1 mirror to the VMFS (instead of having BOTH ESXi AND my VMFS reside on the SATA) then it seems to unmount / sync the LSI cleanly, on a consistent basis.
(This is all purely anecdotal of course. I suspect that when ESXi lives elsewhere, it unmounts the VMFS and syncs it, then initiates shutdown -- it appears to shut down the USB or PERC considerably more cleanly than it seems to shut down and sync the SATA....)
2) If I delete the RAID configuration and use it as a non-mirrored disk, it works fine, syncs, and unmounts fine.
3) Since I went ahead and did THAT, I have the option (from LSI) of flashing RAID firmware, or Initiator-Target firmware, onto the controller. The card has a significant performance increase with the Initiator-Target firmware in place, and this is how I am presently using it.
(................. of course, since it isn't a redundant array, this begets the question: Why the heck don't I just do a crude hack to use the 830's on-board sata?)
TO that point -- yes, I am probably going to get a couple of IDE flash cards and hack ESXi to boot off of those, diskless. If I can't figure out a driver/software solution in the Service Console (this is not mission critical, I could pick up an adaptec 3805 for $250 heh) .... then I might drop this controller and the disks into an iSCSI target running .. well, you say OpenFiler (Linux), I say FreeNAS (FreeBSD) -- but, whatever. At least on the iSCSI target, I can run the appropriate drivers and monitoring utilities to stay on top of the mirror's integrity, as it's exported to ESX.......
re iSCSI: just FYI, I'm crossing over a couple of Compaq SX-1000 (e1000) fiberoptic NICs between ESX and iSCSI specifically for the mount -- it's almost as good as the real thing.
Hope that gives you some info to start off with.
-R
