JonathanM

@craigr Did you solve the issue before you shipped the card?

Most likely yes, but without real numbers of what cards and drives are being powered, who knows?

Multiple containers plus the built in GUI can theoretically share / split access to a video chip. VM passthrough keeps the GPU tied up exclusively. Don't know if that answers your question.

Server grade hardware as long as you can use a plain HBA instead of a RAID controller is typically more reliable in the long term.

Since you don't have parity currently, the easiest path would be to add all the drives, format the new 5TB data drive as unencrypted, then copy the encrypted 4TB to the newly formatted 5TB. Then you can reformat the 4TB to an unencrypted file system. Before you shut down the current system for the last time be sure to disable the VM and Docker systems, and set the array to NOT autostart. That way you can copy any system files and appdata from the current SSD pool to the new NVME when you boot up the new system.

Dashboard includes cpu time spent waiting for i/o, and since preclear is waiting for the disk i/o 100% of the time it's running...

No, it is only relevant for heatsink calculations. It's a measure of maximum allowed heat creation over time, not efficiency. For Unraid, you want the CPU to be able to do the most work possible in the shortest period of time, so all the rest of the hardware can go into low power mode as soon as possible. If you limit the CPU to only use a portion of its ability, you will keep the rest of the system awake and consuming power so your overall consumption for a given job will be higher. As an example, consider transcoding a video file. While the job is running, RAM is active, hard drives are spun up, etc. If you limit the amount of processor power available, it will still require the overall same amount of calculations to finish the file, but take much longer to finish, keeping the consumption of all the other parts higher longer, and still use the same amount of CPU energy, just spread out over more time. Kneecapping your CPU with the T version will cause long term power consumption to be higher on a highly optimized system. Better for overall consumption is limiting the drive count to the absolute minimum, likewise the memory chip count. Also make sure you have enough motherboard SATA ports, addon HBA's are power hogs. 1 16TB disk drive is way more efficient than 4ea 4TB drives.

As long as you don't add or remove array members, parity1 will remain valid regardless of slot assignments. Parity2 is slot sensitive so would need to be rebuilt.

macvlan causes crashes on some systems

P.S. I find it mildly amusing that the UPS cable setting for slave is "ether", and the UPS type is "net" ethernet, get it? Yeah, yeah, small things for small minds.

Slave, note the IP address in the device field, set it to match the fixed IP of your master Unraid. Master, nothing special, as long as the apcupsd is up and running it defaults to broadcasting as master on port 3551 Be sure the network path between the two boxes is powered during an outage, otherwise the slave box will never shut down. I recommend starting shutdown on the slave box almost immediately, so you can be sure it's down cleanly before the master box even starts to shut down. If you have VM's on the two boxes, install the apcupsd program on them as well, use the same slave settings, and start the shutdown first on them. Depending on load, capacity, and desired stress on batteries I recommend starting shutdown on VM's at the 60 second mark, the slave unraid box at 180 seconds, and the master at maximum 300 seconds. That way everything has time for a clean orderly shutdown without distressing the batteries by discharging below 50% or so. Once the heavy draws are shut down, you can leave the UPS running for a little while to keep the network alive. Your situation may differ slightly, but the basic outline will probably be the same.

None. Unraid's parity emulation and rebuilding has nothing to do with the filesystem, it has no concept of files and free space or not, it's just a long string of 1 and 0 across the entire partition.

Try reducing the cores and RAM assigned to the VM. Remember, all the resources assigned to the VM are off limits to the host that is emulating the motherboard and I/O. Slow motherboard with lots of RAM and CPU = poor performance.

Just my opinion, but if your hardware and infrastructure is flawless, BTRFS is probably superior. However, I would characterize BTRFS as fragile, it doesn't recover as well from hardware errors as XFS. ZFS native support is in private beta, but it is very promising as a good alternative to either BTRFS or XFS. If I were in your shoes I'd wait a little while until ZFS is public and see how it goes. Drive encryption is a separate issue. If you are already keeping full up to date backups, then encryption can be a viable option if you want what it adds. However, if you don't have and keep full backups in another location, don't even think about drive encryption. It makes disaster recovery almost impossible comparatively.

Keep in mind that the write performance will be limited by the speed of the parity drive, and trim is disallowed on parity protected data drives, so if that specific model SSD needs regular trimming you could lose read performance as well. Unclear what you mean by this, if both drives are in a single RAID1 pool that is healthy both drives will participate equally.

That's the joke. Similar to duke nukem forever. soon has become the light hearted way of dealing with the seemingly interminable delays between releases. Soon™️ has no time scale attached, it's some date in the future with no way to make a prediction. Even the developers don't have a hard timeline, when it's done is the official answer. 1 month is NOTHING in the historical scale of Soon™️. That said, it could happen any time now. I'm beginning to think the Unraid community is unwittingly participating in a variable interval reinforcement schedule study, https://open.lib.umn.edu/intropsyc/chapter/7-2-changing-behavior-through-reinforcement-and-punishment-operant-conditioning/#stangor-ch07_s02_s02_t01

It's the best way, preferably with static assignments handed out by the router's DHCP service. That way you only have a single place to update things to manage your IP's. However, when the DHCP service is managed by a VM, you are forced into setting it on the host manually. I still have the assignment in the router's DHCP table, as I keep a standalone PC ready to take over the pfSense duties if the server ever needs more than a few minutes downtime. Keeping 2 pfSense instances synced is easy.

Not normal. I run pfSense as a VM, and don't have any of the listed issues. Running your gateway as a VM can cause issues, but not if you have things configured to account for them. In my case, Unraid has a 10GB interface with a fixed IP connected to the main switch, and the pfSense VM has 2 1GB connections passed through, the WAN assigned interface connected to the modem, the LAN assigned interface connected to the same switch as the 10GB. Other than not having DNS and WAN connectivity until the VM is running, Unraid handles it just fine.

Extensively discussed in the support thread. Quick fix is delete nginx.conf ssl.conf and replace them with the corresponding .sample files.

No links without manually searching, sorry. It's always been an issue when someone manually sets the font color, but recently it seems to pop up a whole bunch more, and it felt like something changed vs. suddenly a whole bunch of people decided to set their fonts.

Try looking through the support thread and if your questions aren't covered ask there.

BTW, I wouldn't call this an issue, more of a hmm, that's interesting. Reason being, board manufacturers spend most of their resources making sure ms windows works, very little effort is put into alternative OS's. Just because the board doesn't detect the OS loading sequence doesn't mean it's not working. Probably a BIOS update would be needed to fix it, but I doubt the board manufactures even know or care, so it's probably not going to get fixed. It's not something that Unraid is going to address.

Same behaviour in GUI or console boot mode?

No, the generated app password.

Another post example.