So, I decided to build a big old OI based storage mule. I started with a HP DL380G6, dual cpus, 80GB RAM, and the onboard RAID driving it's 2.5" SAS cluster as a RAID 5 for the OS. Not ideal but the HP RAID controller doesn't do IT mode. I may break it out into dual RAID 5s and let OI mirror them for extra piece of mind...
For storage control I got a LSI 9207-8e. This would feed an Addonics RR2035RSDES shelf with 20 bays. To keep things snappy I also sourced three Intel DC3600 400GB PCIe SSDs, two to be mirrored as a log volume, one for cache. I've got an existing fiberchanel fed farm that I want to tie into so I've tossed in a dual 4GB port FC card. From there it'd be a happy time just filling it up with disks. I opted to start with five WD Red 6TB units as a test.
My research showed the LSI dealt with multiple disks off a given port, and I knew SAS controllers would talk to SATA so I thought all was good. I failed to discover that 'SATA port multipliers' and 'SAS port multipliers' are not treated equally. Net result at most I could get four drives recognized, one off each SATA PM in the chassis. Research fail on my part.
I'm now stuck trying to massage this collection of bits back into usable service for as low a cost as possible. As the Addonics case is JUST a drive enclosure I'm limited to a SAS expander that can be powered by a molex connector or similar, there isn't a PCIe slot in there. That naturally has led me to the Chenbro CK23601. Ideally I'd like to go SFF-8088 to SFF-8088 from my LSI to the CK, possibly dual linking. I'll be able to mount the CK using a plexi-adapter sheet in the case to accommodate this nicely. The docs and Chenbro support say no, they insist you must use the one designated SFF-8087 port as your input from the HBA. I've seen user reports that claim otherwise in this forum, but nothing recent.
Anyone care to post a current update on the CK and how flexible it is on port designation?
I've also got to chase why OI only sees two out of three NVMe SSDs but that's a problem I can solve down the road.
For storage control I got a LSI 9207-8e. This would feed an Addonics RR2035RSDES shelf with 20 bays. To keep things snappy I also sourced three Intel DC3600 400GB PCIe SSDs, two to be mirrored as a log volume, one for cache. I've got an existing fiberchanel fed farm that I want to tie into so I've tossed in a dual 4GB port FC card. From there it'd be a happy time just filling it up with disks. I opted to start with five WD Red 6TB units as a test.
My research showed the LSI dealt with multiple disks off a given port, and I knew SAS controllers would talk to SATA so I thought all was good. I failed to discover that 'SATA port multipliers' and 'SAS port multipliers' are not treated equally. Net result at most I could get four drives recognized, one off each SATA PM in the chassis. Research fail on my part.
I'm now stuck trying to massage this collection of bits back into usable service for as low a cost as possible. As the Addonics case is JUST a drive enclosure I'm limited to a SAS expander that can be powered by a molex connector or similar, there isn't a PCIe slot in there. That naturally has led me to the Chenbro CK23601. Ideally I'd like to go SFF-8088 to SFF-8088 from my LSI to the CK, possibly dual linking. I'll be able to mount the CK using a plexi-adapter sheet in the case to accommodate this nicely. The docs and Chenbro support say no, they insist you must use the one designated SFF-8087 port as your input from the HBA. I've seen user reports that claim otherwise in this forum, but nothing recent.
Anyone care to post a current update on the CK and how flexible it is on port designation?
I've also got to chase why OI only sees two out of three NVMe SSDs but that's a problem I can solve down the road.