My first experience installing an LSI HBA yesterday was borderline catastrophic and I wanted to get some help diagnosing the problem before I make another attempt, as I nearly lost my entire primary pool in the process. (Sorry for the length, the situation is a bit complicated.)
To set the stage, I picked up an HP H220/9207-8i pre-flashed to IT mode from Art of Server on Ebay, and a pair of these cables. My FreeNAS box has an encrypted (yes I know) RAIDZ1 pool, 4x6TB previously connected to the motherboard's SATA controller.
Installed the card, connected the four drives to one of the HBA's ports, and because I wanted to test both ports and cables to double-check that everything was good, also connected four old SSDs to the HBA's other port.
On first boot, the HBA's BIOS screen sat at the "Initializing... \" animation for a very long time, what felt like 5-ish minutes before dumping its report of connected drives faster than I had a chance to read and finally booting the machine. (Note I didn't enter the configuration utility here, but my impression per Art of Server is the card should have essentially been JBOD out of the box as he handles all the flashing before shipment.)
The FreeNAS boot threw a ton of hardware/read errors, and when I got to the web UI, sure enough only three of the hard drives and three of the SSDs were visible (putting the pool into a degraded state, obviously). To try to rule out a bad cable, I shut down and swapped two of the SATA connectors between the hard drive that failed to connect and one of the good ones. Powering back on, the drives had swapped places: the previously good drive failed to register, and the one that had originally failed to connect was now connected -- but it had become UNAVAIL per zpool status.
At this point I panicked, pulled the HBA, and reconnected the four drives to the motherboard's SATA ports. In some stroke of luck or divine intervention, three of the drives did come back up as ONLINE, but the fourth remained UNAVAIL. If a second drive had gone UNAVAIL this would be a much sadder and angrier post, as I didn't have a backup of this pool (in my defense, setting up a permanent backup is the end goal of getting this HBA in the first place). I was unable to easily "zpool online" the unavailable drive, so before I risked further banging on a degraded array, I grabbed a pair of big USB hard drives and am currently backing up the entire pool. (Rather than trying to zpool replace the corrupted drive after the backup is done, I may take this opportunity to simply destroy the pool and recreate it unencrypted and restore the data, since running an encrypted pool has been a lot of hassle and worry for little benefit so far.)
Does anyone have any insight into this behavior with the HBA? Did I make any obvious mistakes in the installation process? The results of swapping the SATA connectors would suggest that the cable is at fault, but it's hard to believe that both cables in the package are bad in the exact same way, and further, I'm confused as to why a bad cable could have been so destructive to one of my drives. I'm not sure how to proceed here, whether to order another set of cables or contact the HBA seller, though I will reinstall the card and continue troubleshooting once I'm done backing up the pool and there's no risk of total data loss.
To set the stage, I picked up an HP H220/9207-8i pre-flashed to IT mode from Art of Server on Ebay, and a pair of these cables. My FreeNAS box has an encrypted (yes I know) RAIDZ1 pool, 4x6TB previously connected to the motherboard's SATA controller.
Installed the card, connected the four drives to one of the HBA's ports, and because I wanted to test both ports and cables to double-check that everything was good, also connected four old SSDs to the HBA's other port.
On first boot, the HBA's BIOS screen sat at the "Initializing... \" animation for a very long time, what felt like 5-ish minutes before dumping its report of connected drives faster than I had a chance to read and finally booting the machine. (Note I didn't enter the configuration utility here, but my impression per Art of Server is the card should have essentially been JBOD out of the box as he handles all the flashing before shipment.)
The FreeNAS boot threw a ton of hardware/read errors, and when I got to the web UI, sure enough only three of the hard drives and three of the SSDs were visible (putting the pool into a degraded state, obviously). To try to rule out a bad cable, I shut down and swapped two of the SATA connectors between the hard drive that failed to connect and one of the good ones. Powering back on, the drives had swapped places: the previously good drive failed to register, and the one that had originally failed to connect was now connected -- but it had become UNAVAIL per zpool status.
At this point I panicked, pulled the HBA, and reconnected the four drives to the motherboard's SATA ports. In some stroke of luck or divine intervention, three of the drives did come back up as ONLINE, but the fourth remained UNAVAIL. If a second drive had gone UNAVAIL this would be a much sadder and angrier post, as I didn't have a backup of this pool (in my defense, setting up a permanent backup is the end goal of getting this HBA in the first place). I was unable to easily "zpool online" the unavailable drive, so before I risked further banging on a degraded array, I grabbed a pair of big USB hard drives and am currently backing up the entire pool. (Rather than trying to zpool replace the corrupted drive after the backup is done, I may take this opportunity to simply destroy the pool and recreate it unencrypted and restore the data, since running an encrypted pool has been a lot of hassle and worry for little benefit so far.)
Does anyone have any insight into this behavior with the HBA? Did I make any obvious mistakes in the installation process? The results of swapping the SATA connectors would suggest that the cable is at fault, but it's hard to believe that both cables in the package are bad in the exact same way, and further, I'm confused as to why a bad cable could have been so destructive to one of my drives. I'm not sure how to proceed here, whether to order another set of cables or contact the HBA seller, though I will reinstall the card and continue troubleshooting once I'm done backing up the pool and there's no risk of total data loss.