-Daedalus

I did, they came back ok. Have been running fine since. I have replacement SSDs on the way anyway (yay Prime Day...). Will keep an eye on it.

Hi all, Server crashed at some point after sleep last night. Didn't notice until I got back this evening. I came back to a server online, but both my M.2 SSDs were missing. These were recently installed in an M.2 PCIE adapter to cool them a little better. They also have a fair amount of medium errors racked up from when they were overheating when first installed. All of this is to say there are absolutely hardware causes for this. I'm just curious if someone spots something obvious in the logs that I've overlooked that could also be a culprit. The plan is to go to 6.12, but I'm going to let it mature a little first. server-diagnostics-20230712-1732.zip

Nice! Scheduling of snapshots?

Or an option with Mover Tuning to ignore settings and move certain shares on a schedule or something. I have the same issue; I have my backups going to an SSD for write speed, but I don't want them sitting there taking up space that could otherwise be used for frequently accessed files that I'd like to keep on the SSDs for a longer period.

In order of preference: Native backup solution (for pools and array to other pool/array, unRAID server, or network location). There's a plug-in, I know, but something more granular than a scheduled zip would be nice. VM backup and snapshots from GUI A way to remove a disk (array shrink) without putting data at risk (at no point should parity be rebuilding without redundancy, or invalid).

Updated. I'll keep an eye on it and report back. Cheers.

Thanks for the quick response! They were overheating originally, so I added some heatsinks, though my understanding is that the SSDs with throttle performance to avoid errors just like this, so I'm not quite sure what's going on.

Google-fu not great here. BTRFS and what look like errors from the controller from the drive (Samsung 970 Evo Plus). It's in a 2 drive mirror. I wonder if someone who's better with BTRFS can have a look, or point me in the right direction. Scrutiny is showing a ton of medium errors, but it's been doing this on both drives, from day 1. I updated firmware at the time, but it didn't seem to have any affect, and the drives have been running fine for well over a year at this point. Snippet: Apr 26 07:03:37 server kernel: BTRFS warning (device nvme1n1p1): i/o error at logical 812252712960 on dev /dev/nvme1n1p1, physical 469695516672, root 5, inode 1072832, offset 4437626880, length 4096, links 1 (path: domains/eco/vdisk1.img) Apr 26 07:03:37 server kernel: BTRFS error (device nvme1n1p1): bdev /dev/nvme1n1p1 errs: wr 0, rd 21284, flush 0, corrupt 0, gen 0 Apr 26 07:03:37 server kernel: BTRFS warning (device nvme1n1p1): i/o error at logical 812249567232 on dev /dev/nvme1n1p1, physical 469692370944, root 5, inode 1072832, offset 4434481152, length 4096, links 1 (path: domains/eco/vdisk1.img) Apr 26 07:03:37 server kernel: BTRFS error (device nvme1n1p1): bdev /dev/nvme1n1p1 errs: wr 0, rd 21285, flush 0, corrupt 0, gen 0 Apr 26 07:03:41 server kernel: nvme_log_error: 22 callbacks suppressed Apr 26 07:03:41 server kernel: nvme1n1: I/O Cmd(0x2) @ LBA 917376120, 8 blocks, I/O Error (sct 0x2 / sc 0x81) MORE DNR Apr 26 07:03:41 server kernel: blk_print_req_error: 22 callbacks suppressed Apr 26 07:03:41 server kernel: critical medium error, dev nvme1n1, sector 917376120 op 0x0:(READ) flags 0x800 phys_seg 1 prio class 0 Apr 26 07:03:41 server kernel: scrub_handle_errored_block: 14 callbacks suppressed Apr 26 07:03:41 server kernel: BTRFS warning (device nvme1n1p1): i/o error at logical 812252721152 on dev /dev/nvme1n1p1, physical 469695524864, root 5, inode 1072832, offset 4437635072, length 4096, links 1 (path: domains/eco/vdisk1.img) Apr 26 07:03:41 server kernel: nvme1n1: I/O Cmd(0x2) @ LBA 917369976, 8 blocks, I/O Error (sct 0x2 / sc 0x81) MORE DNR Apr 26 07:03:41 server kernel: critical medium error, dev nvme1n1, sector 917369976 op 0x0:(READ) flags 0x800 phys_seg 1 prio class 0 Apr 26 07:03:41 server kernel: btrfs_dev_stat_print_on_error: 14 callbacks suppressed Apr 26 07:03:41 server kernel: BTRFS error (device nvme1n1p1): bdev /dev/nvme1n1p1 errs: wr 0, rd 21300, flush 0, corrupt 0, gen 0 Apr 26 07:03:41 server kernel: nvme1n1: I/O Cmd(0x2) @ LBA 917379192, 8 blocks, I/O Error (sct 0x2 / sc 0x81) MORE DNR Apr 26 07:03:41 server kernel: critical medium error, dev nvme1n1, sector 917379192 op 0x0:(READ) flags 0x800 phys_seg 1 prio class 0 Apr 26 07:03:41 server kernel: BTRFS warning (device nvme1n1p1): i/o error at logical 812249575424 on dev /dev/nvme1n1p1, physical 469692379136, root 5, inode 1072832, offset 4434489344, length 4096, links 1 (path: domains/eco/vdisk1.img) Apr 26 07:03:41 server kernel: BTRFS error (device nvme1n1p1): bdev /dev/nvme1n1p1 errs: wr 0, rd 21301, flush 0, corrupt 0, gen 0 Apr 26 07:03:41 server kernel: BTRFS warning (device nvme1n1p1): i/o error at logical 812254294016 on dev /dev/nvme1n1p1, physical 469697097728, root 5, inode 1072832, offset 4439207936, length 4096, links 1 (path: domains/eco/vdisk1.img) Apr 26 07:03:41 server kernel: BTRFS error (device nvme1n1p1): bdev /dev/nvme1n1p1 errs: wr 0, rd 21302, flush 0, corrupt 0, gen 0 Apr 26 07:03:41 server kernel: nvme1n1: I/O Cmd(0x2) @ LBA 917382264, 8 blocks, I/O Error (sct 0x2 / sc 0x81) MORE DNR Apr 26 07:03:41 server kernel: critical medium error, dev nvme1n1, sector 917382264 op 0x0:(READ) flags 0x800 phys_seg 1 prio class 0 Apr 26 07:03:41 server kernel: nvme1n1: I/O Cmd(0x2) @ LBA 917385336, 8 blocks, I/O Error (sct 0x2 / sc 0x81) MORE DNR Apr 26 07:03:41 server kernel: critical medium error, dev nvme1n1, sector 917385336 op 0x0:(READ) flags 0x800 phys_seg 1 prio class 0 Apr 26 07:03:41 server kernel: BTRFS warning (device nvme1n1p1): i/o error at logical 812255866880 on dev /dev/nvme1n1p1, physical 469698670592, root 5, inode 1072832, offset 4440780800, length 4096, links 1 (path: domains/eco/vdisk1.img) Apr 26 07:03:41 server kernel: BTRFS error (device nvme1n1p1): bdev /dev/nvme1n1p1 errs: wr 0, rd 21303, flush 0, corrupt 0, gen 0 Apr 26 07:03:41 server kernel: nvme1n1: I/O Cmd(0x2) @ LBA 917388408, 8 blocks, I/O Error (sct 0x2 / sc 0x81) MORE DNR Apr 26 07:03:41 server kernel: critical medium error, dev nvme1n1, sector 917388408 op 0x0:(READ) flags 0x800 phys_seg 1 prio class 0 Apr 26 07:03:41 server kernel: nvme1n1: I/O Cmd(0x2) @ LBA 917373048, 8 blocks, I/O Error (sct 0x2 / sc 0x81) MORE DNR Apr 26 07:03:41 server kernel: critical medium error, dev nvme1n1, sector 917373048 op 0x0:(READ) flags 0x800 phys_seg 1 prio class 0 Diags attached as well. server-diagnostics-20230506-1116.zip

Well I'm looking at it because I'm maxing out CPU some of the time (4K/HDR/HEVC library) with enough users, so I'd be upgrading for perf reasons anyway, I'm just thinking to reduce power while I'm at it, given that having that CPU stuck at 100% for hours at a time doesn't feel great. I hear you on ROI. My server at the moment is costing around €35-40 per month to run, so if I could cut idle power by 50W and lots-of-plex-power by maybe 100W(?), the savings there would be enough for a couple of coffees a month, at least. Larger drives would be nice, but I've been continually expanding the server as the content grows, so I'll just be phasing out the 8TB drives with 18/20TB ones as needed.

Thanks for the response. That's significantly lower than mine alright. I know the HBAs and NIC will be pulling a certain amount of power (and probably the backplane as well, come to think of it), but my idle draw still seems high, though I suppose that's what happens with a 4-year-old architecture at this point. I'd like to go intel for iGPU, and I like the idea of having game server stuff on performance cores, and unRAID stuff on efficiency cores. Only downside is LGA 1700 server boards are expensive, and there aren't many consumer boards that offer the layout or features that I'm after.

Hi all, Electricicy getting more expensive these days, I'm taking a closer look at my server's current power usage, and eyeing up a more power efficient build. At the moment, my server idles at 140W. All drives spun down, nothing running. 170W with drives spinning. Hardware: 13 8-18TB drives, 2 SATA SSDs, 2 NVME drives, 2 9207-8i controllers, an X-540-T2 NIC, and a 3900X, running on an X470D4U The controllers use around 10W each, and the NIC can use around 15-20W. Let's assume another 5W each for the solid state storage, and that means my CPU is pulling around 80W at pretty much zero load, which seems high. I have turbo enabled, CPU governor set to power save. Can anyone with anything similar share numbers, in case I might have missed something here? Second, I'm thinking of switching to something like an i5 13500, for Plex to use the iGPU rather than CPU at the moment. What are people's thoughts on this? I assume power consumption is pretty low while transcoding but it's a difficult thing to find numbers on. Anyone have a better idea?

Makes sense actually, thinking about it. With ZFS being added, and with the plan to make an "unRAID" array just another type of pool, it's very likely people will use unRAID without a traditional array, so shouldn't have the overhead at all.

+1, definitely. I believe this is done for VM vdisks for the same reason, would be nice to at least have the option for shares too.

Not at the moment. Your best bet is to create a new pool with the extra disks. If you're on the stable version of unRAID, this will be a BTRFS pool, defaulting to RAID1, but this can be changed to RAID5/6 if desired. If you're using a beta version of 6.12, you can also choose a ZFS pool.

I'm gonna go out on a limb and say there will be more than one. 😋

Thanks for the episode! Really good listen! I should have made my original question clearer: I more meant, why aren't certain plugins installed by default. If they were OS-native that is a different story.

I hear you, and I kind of agree to be honest; I don't see a phone as a common interface for the unRAID UI. I was more making a point that the "It works fine for me" type responses from a prominent community member/unRAID developer don't generally sit well with people (as I'm sure you guessed)

The request was for a mobile-friendly UI. The confirmation of ability to use the current UI on a single device does not answer the request in any way.

Squid, as an author of many widely-used plug-ins, which (if any) do you think should be shipped as part of the OS proper? Can you talk about why this isn't the case? Same question to Ed.

+1 for this too. I'm dumping stuff into syslog at the moment, but UI pop-ups would be sweet.

LSI 9207-8i is a good bet. It's PCIE 3.0, so you'll get ~500MB/s to each drive, if you want SATA SSDs in there at any point, and it's only a HBA, so no flashing needed. Or if you know you're only ever going to use HDDs with it, then a 9211-8i will do fine (you'll get around 330Mb/s max per drive)

Sounds like it'll be a more direct competitor to the QNAPs/Synologys out there. Interesting times. unRAID is amazing in its flexibility, but (I think) it's lacking in the polish department. Things like downsizing or removing disks from the array, rebuilds (without having to remove parity first), a native backup solution, and general system monitoring, to name a few. It will be interesting to see where the focus is for this new product, and to see how they handle the more complicated stuff (if at all) that unRAID offers.

Second example looks spot on. For the first one, I believe rsync always expects a directory, but I think if you change the desination to "/mnt/disk1/backup/fileio.img" it should be alright. I'd advise running the above with the "--dry-run" option, so that you can verify what will happen. If you're happy, remove that and kick it off.

Yeah sure, my bad! #!/bin/bash logger "SSD Pool > Array backup started" The inner-workings of very first line aren't super-important, just know that it should be at the start of every script you make for now. This first bit just sends a message to the unRAID syslog "SSD Pool...." # Backup things from SSD pool to array disks rsync -avh --del --stats \ This next bit is just a comment, saying what it's doing. The next line is the start of the rsync command. In general, it looks like: rsync -someOptionsHere /path/to/source/directory /path/to/destination/directory. So with this one, the options are a, v, and h. a = archive. Basically what you want. Copy any new files from source, to destination, overwriting if the file is newer, but don't delete anything from either side. Also keep all the ownership, permissions, etc. v = verbose. Just log what it's actually doing while it's doing it, in case something screws up and I need to look at it later. h = human-readable. Just so it's in KB/MB/GB instead of bytes. Then the next bits are more options, but they don't have single-letter shorthand that can be combined together, so you have to specify them separately. --del = delete from source directory after copy to destination is finished. Because the whole point is we want stuff off the cache here. --stats = more info about what it did at the end of the operation, again for logs. The very last thing is the \ symbol. This basically says, continue the command, but on the next line. Just for my own readability. --exclude 'backup/' \ --exclude 'downloads/' \ --exclude 'movies/' \ --exclude 'tv/' \ --log-file=/mnt/user/backup/ssd-pool/backup-log.txt \ /mnt/ssd/ /mnt/user0/backup/ssd-pool/ These continue the options like before, this time to exclude certain directories from the copy, becuase I don't care about them much, so it's not the end of the world if they get lost. I mostly made this script to keep my documents and pictures a little more safe. The second-last line is an option to write all the logging we've been doing to a file. In this case "backup-log.txt" on the SSD. And the very last line specifies our source and destination. So my cache is called "ssd" so we have "/mnt/ssd" as the source, and I want it to go to a share called "backup" and a folder in there called "ssd-pool". The important bit here though is that we don't want it going to the user share (/mnt/user/backup) because this includes the cache and the spinning disks, so the files wouldn't actually move, or might even get corrupted, so we specify "user0" (this is a short-hand for "the share, but exclude cache"). Hopefully all that made sense!

This is what I meant. Sure. You'll need to change the names of shares obviously, and I take no responsibility for your server catching fire, etc. #!/bin/bash logger "SSD Pool > Array backup started" # Backup things from SSD pool to array disks rsync -avh --del --stats \ --exclude 'backup/' \ --exclude 'downloads/' \ --exclude 'movies/' \ --exclude 'tv/' \ --log-file=/mnt/user/backup/ssd-pool/backup-log.txt \ /mnt/ssd/ /mnt/user0/backup/ssd-pool/ logger "SSD Pool > Array backup finished"