So, this past Friday night I lucked out as I saw the three of my kids that were home, actually go to bed by 8pm – and even fall asleep. I had much geeking to do. Very much.
I had already put most of my new components in the case when I got home from work. I did that while I was in the kitchen, as my wife was getting to leave for work. It would just be less for me to do later in case I ran into any issues.
The Norco RPC-470 envelops the mATX-sized Intel BOXDH61BEB3. If price hadn’t been such a factor, I would have gone with something smaller than this, but at least I have good airflow. I don’t have the power lines tied up nicely as I still have to put in the drive cages at this point.
Once the kids were asleep and my wife was long gone to work, I took it to the basement to begin the real work. I shut down my old unRAID hardware for the last time and yanked it out of service. The next step was to get the SuperMicro AOC-SASLP-MV8 and Koutech USB header adapter installed with the unRAID flash drive.
Now it was time for the drive cages. The first one was coming out of an external enclosure that implemented my port multiplier and had a 5-in-3 drive cage with 5 hotswap drives.
The 5-in-3 cage went into the left-hand emplacement and was screwed down in-place from the top.
Next was the 4-in3 cage, which was installed internally on the unRAID server itself.
Installing the bay into the middle drive bay area required me to use the included HDD brackets as there were no mounting holes to screw it into place.
Using those, it slid in and locked into place.
I now have room for an additional 5-in-3 cage so that I can expand to 14 hot-swappable drives if needed. All that was left at this stage was to hook up the power and SATA cables and try to tidy it up a bit. It’s not perfect, but it’s good enough.
Now we have to make sure this thing will boot. I had tested it out prior to all of the prior just to make sure all the components were good and that test passed. Now it was time to make sure all the drives would be detected and to make some last minute BIOS adjustments.
I also disabled unused integrated devices. I don’t see the point in using the interrupt if it’s not being used. Active unused devices just use memory and can cause unneeded issues.
After a few other minor adjustments, I saved the BIOS settings and let the system continue booting. This is the first good sign – the unRAID flash did boot.
Further success – the logon prompt.
Unfortunately, this was the last good news I would get for the next few hours. Immediately upon getting this logon prompt, I switched over to a Windows system so that I could access the webGUI – it was a no go. Unable to connect. I also found I was unable to ping. Anyway, long story short, I found after much digging that the Intel 82579V integrated NIC chipset apparently has no available driver in the final 4.7 unRAID release. Therefore, I had no network connection at all. I confirmed this by running ifconfig from command line and was only given statistics for the loopback adapter. Suck.
So, I now have two options:
- Revert back to at least the old motherboard, CPU, and memory.
- Upgrade unRAID to 5.0rc8a (latest release candidate at the time).
Well, I bought this stuff and I’m not much for retreating, so I decided to soldier on with 5.0rc8a. I had been running an early free beta of 5.0 on a dedicated file server and it has been working fine for around a year now, so I figured, “what the hell?”
All that was required was to get the flash, backup the entire contents, and then copy the bzimage and bzroot files from the 5.0 version over the existing ones (there are some additional steps if you use user shares with users other than root). So, I did that, reinstalled my flash drive, and booted again. I logged in once it made it back to the prompt and immediately ran another ifconfig – success!
Now I could go and access the webGUI and begin reconfiguring my array to start. As you can see, the new 5.0 landing page with further instructions.
Now that I have a working server, I have to get my drives back into the order they were originally. I know this because before I took my unRAID server down, I saved a screenshot of my devices page.
Upon accessing the 5.0 Main page, the array will be stopped and all available disks will be available to change the order with via dropdown boxes. I just made changes to those that were not correct and I was all set. All drives are green and my configuration is valid.
Time to start the array and see what happens.
All is well with the array – started with no issues. However, I’m not done. unRAID 5.0 requires a permissions script to be run, so I have to head over to the Utils tab and go to New Permissions. I check the box to ensure I want to do this and click the Start button. For a collection of my size, this process took about 20-25 minutes. The screen refreshes periodically to inform you of the status, disk by disk.
Once this completes we’re all done…with the unRAID server (unless you need to recreate your users if you use user secruity – I do not). There is one final step that MUST be done upon successfully completing the permissions script. All Windows systems that access the server have to be rebooted. Otherwise, you’ll find yourself facing massive unRAID rage quit. Due to the way credentialing works between Windows and the unRAID server, the permissions script makes Windows fail on my disk and user shares after it completes. Rebooting clears the Windows credential cache and access is restored.
After rebooting all of my systems, I no longer encountered this error, so it definitely works. I was now ready to test some copying to my unRAID server…just to make sure I still had write permissions. That panned out fine and with transfer rates that I was used to expecting.
Those results are to an older, near-full disk, so at around 25MB/s would be correct. Maybe even a little better than what I normally would get. Time to go watch some movies and let this thing do what it does.
After two days of service, I have seen no ill-effects of upgrading either the hardware, or moving to the 5.0rc8a version. I have not run a parity check to see what kind of performance gain I will get there, but it can’t be worse than what I was experiencing with the port multiplier in place. I will run one this week and update this post with my comparison. I was never able to get more than about 25MB/s when the check was running over the port multiplier. It would kick up to around 27-28MB/s when it hit drives connected directly to the board, but those also happen to be my oldest (slowest) drives. We’ll see!
So, I said I’d update when I was able to run a parity check and I just did. Wow. Not much else to say other than that. My old hardware was only able to max out at around 32MB/s, but I would say it averaged about 27MB/s. The numbers below say it all.