Recent content by MilhouseVH

Fair enough, and you've been fortunate, but it's still a potential lottery that can be entirely avoided by using software RAID which really imposes so little overhead these days. Hardware RAID made more sense 10 or so years ago when CPU and IO performance was both low and expensive.

Hardware RAID is definitely overkill for this box, and it's not even something I'd recommend for more powerful and more expensive solutions since hardware RAID forces you to replace failed controllers with identical replacements (good luck doing so in 5-10 years time, maybe find one on eBay...)...

I would avoid any controller with vertical ports, or that has ports at the front by the IO plate for the very reasons you mention - card installation and cable management becomes a total nightmare with the lack of space due to the PSU being mounted above the card which will exert pressure on one...

Edit: Ooops - just realised I posted this to entirely the wrong forum! :) To fill in the blanks - the fan in my N36L PSU began making grinding noises after 15 months so I searched for a replacement. The HP original PSU fan is a T&T 4020HH12S-ND1 (40x40x20mm 12V DC sleeve bearing, 7000RPM...

@zero - your links are all broken (I'm not likely able to offer you any additional assistance but at least thought I'd let you know about your links!)

I was under the impression that a single disk ZIL was not considered a good idea, and that best practice is to use a mirrored ZIL because - for pool versions prior to v19 - if you lose your ZIL you may lose your entire array.

Old thread revival, but I just found this interesting post and although it relates to SSDs, it claims that the B5/181 "Program_Fail_Cnt_Total" SMART attribute - in a Crucial SSD - is an indication of 4K mapping (in)effectiveness, it's the number of unaligned reads and writes. Could this also be...

The rear system fan slides into a snug fitting frame, so I doubt there's any room for grommets. Maybe the PSU fan could be quietened somehow, but when operating normally these boxes are almost totally silent - it's only when they start buzzing and vibrating for no apparent reason that they can...

This is correct, only the newer 6Gb/s LSI controllers support 3TB+ - the older 3Gb/s adapters are limited to 2TB.

HP Microserver N40L, with AMD Turion II Model Neo CPU and 2GB RAM for $20/£10 more than the N36L. It seems the new CPU might be clocked at 1.5GHz rather than the 1.3GHz of the older AMD Athlon Neo in the N36L. Not a major improvement, but for only $20/£10 the extra speed might come in handy...

I suppose you could try the latest BIOS released Aug 2011, I'm going to try it now with the LSI 9211. Edit: Didn't work for me - no change.

I don't think the BIOS in the HP Microserver leaves enough "room" for the LSI BIOS to load - it's a common problem. I also get the same problem with an LSI 9211-8i in the N36L, but fortunately having flashed the 9211-8i with IT (Initiator Target) firmware I've got no reason to access the LSI...

Exactly, both vdevs are RAIDZ1, meaning I can lose up to 1 disk from each vdev before losing the whole array/zpool - this is adequate redundancy for me as I also have backups on a separate server. Unless you really need the performance, I think there's a lot to commend 5400RPM drives.

The 5400RPM 2TB disks are nice and quiet, cheap, and do the job! :) The 500GB drives just happened to be what I had lying around... I also configured another N36L for my brother with just the standard onboard SATA2 and another Intel CT NIC (the onboard Broadcom is totally busted in FreeBSD...

Merely future proofing - whether I actually bother with SATA3 disks in the future depends on my bank balance at the time. :) The 9211-8i is PCIe v2 8x which means it can sustain more than adequate throughput for 8 disks (500Mbytes/sec each lane), whether SATA2 or 3. The SAS3081E-R is PCIe...