jbartlett

I think I figured out what happened. In the Historical section, it lists: Samsung_SSD_960_EVO_500GB_S3X4NB0K309824V (Samsung_960EVO) In the UD list, it has: Samsung SSD 960 EVO 500GB_S3X4NB0K309824V + 1: Samsung_960EVO It looks like rc8 removed the underscores from the drive ID so it didn't find a match and created it as if it was new.

I updated all plugins and then upgraded from 6.9.2 to 6.10.0-rc8. The partition names identified by UD changed back to the default serial number and not the value I had set previously and the 6.9.2 version was listed in the "Historical Devices" section. Reverting back to 6.9.2 restored the partition name. Under 6.9.2: Samsung_SSD_960_EVO_500GB_S3X4NB0K309824V (Samsung_960EVO) Once I renamed the partition back to "Samsung_960EVO", it fixed the mounting location so my VM's could be located again. nas-diagnostics-20220511-2047 (rc8).zip nas-diagnostics-20220511-2052 (6.9.2).zip

The container port must remain 8888.

Make sure your antivirus, Acronis Home, etc are disabled and ensure you have the prompt to run as Admin show up.

I kept getting data corruption and missing files and a parity sync kept reporting errors if I wait a few days after the last parity check. I managed to isolate it to one drive I swapped it out and did a parity rebuild but the issue persisted so I assumed a corrupted XFS. I was using this utility to help track corruptions and it was a clear benefit there. It seemed that removing all files fixed whatever corruption was happening, had no issues even after filling up the drive with large NULL files in the event I needed to 7zip the drive up to give someone to check but it never failed in the next month. Since I needed to reformat, I decided to switch to btrfs to use it's own built in scrub for bitrot, and I was thinking of taking advantage of btrfs features.

One thing to note is that the memtest86 that comes with UNRAID is an old version* - if you want the most recent, you have to download it from https://www.memtest86.com/ and create your own boot flash. Note that this version restriction placed by memtest86. * At least this was the case the last time I used the tool, YMMV

I've been pondering software KVM's off & on and looked into Barrier. It strikes me that if you create a VM running Windows or Unix, you can use that as your Barrier host.

Sometime last week, I set my Validation scheduled to weekly on Sunday. Disk 1 kicked off and completed but none of the others did. I'm also wondering how does the Task function work. You can set up One drive per Task which seems the logical choice but you can also assign multiple drives to a single Task. What exactly *is* as "Task" in this case? It seems like a single thread that will do stuff. There's no built-in help around it.

Seems like it.

When ever I tried to schedule one to kick off daily in 15 minutes, it's only ever kicked off the check for Drive 3. It finished and nothing else started. *shrug*

A workaround is to export the drive and then check the export on that drive.

The "Show Problems" link doesn't show the problems. Clicking the link opens a popup dialog with no contents. The file /var/tmp/disk1.tmp.end does exist.

It'll report something like "1 mismatch (updated)" or "1 corruptions"

Feature request: Export by share so hashes from different machines can be compared against each other without manipulating the export files. For example, comparing files between a primary server and a backup server with different drive layouts.

I've gotten it to run under a different unix OS (I forget which) but the main thing you need to ensure is Privileged=true. This is also the main reason why DiskSpeed doesn't work in Windows because it doesn't support that flag. From there, it's just experimentation.

Search your syslog file for "recovery thread". It'll list block numbers that failed. Issue is you really don't know which drive the fail came from. I recently run into this issue with one of my drives and currently adding support to the version 3.0 of DiskSpeed to search your drives to find what files occupy that spot on all of the drives for further checking. If you have 3+ errors (I had 100+), look for a pattern such as 1,4,8,12 where it skips 3 or some number each time - that's a sign that one side of a platter is failing. It could be on your data drives, it could be on your parity drive. I'd also recommend getting the "Dynamix File Integrity" plugin.

The only way I can see how to invoke a scan of a drive is to go to the app's settings and change the cron job and drive settings to kick off the validate process in the next available time slot.

Amazon isn't doing any counterfeiting. Someone sold fake stuff to Amazon and Amazon didn't catch it. For their part, Amazon evidently doesn't keep the same (apparent) stock from different sources separate.

I've been working on adding support for benchmarking SSD's and wanted to post my method for others to weigh in on. The method I came up with allows DiskSpeed to come up with the following metrics using a Western Digital Black NVMe 250GB, model WDS256G1X0C: Read Speed: 427 MB/s Write Burst Speed: 202 MB/s (ends after 7.52GB) Write Sustained Speed: 118 MB/s - 149 MB/s There needs to be at least one partition mounted with at least 15GB of free space available. In this example, I will use "/mnt/cache". SSD's in a RAID or drive pool will not be tested as it is not possible to test just one drive in the series. The commands given below are formatted for repeating on your personal setups. For best results, shut down any VM's or active Dockers residing on the cache drive. 1. Sync drive sync "/mnt/cache" 2. Trim drive fstrim -v "/mnt/cache" 3. Create 10GB file with random data dd if=/dev/random of="/mnt/cache/DiskSpeed_fq9.junk" bs=1MB count=10000 conv=noerror status=progress 2> /boot/cache_write.txt 4. Sync drive sync "/mnt/cache" 5. Read the file dd if="/mnt/cache/DiskSpeed_fq9.junk" of=/dev/null iflag=direct status=progress 2> /boot/cache_read.txt 6. Delete the file & trim the drive again The line delimiter used for cache_read.txt & cache_write.txt is just a carriage return so you will need to use a program that can understand that such as Notepad++ or it will all appear on one line. The last line in cache_read.txt is used for the read speed. 10000000000 bytes (10 GB, 9.3 GiB) copied, 23.4231 s, 427 MB/s The cache_write.txt file is read line by line looking at the number of bytes read, the current duration, and speed. 203000000 bytes (203 MB, 194 MiB) copied, 1.00453 s, 202 MB/s When the number of seconds from the last line compared to the current line is more than 2 seconds, then the drive's buffer was filled and there is a pause while the drive flushed some of buffer to make room. The highest value taken up to this point is considered the Write Burst Speed as it utilizes the drive cache memory. The number of bytes at the current line (gap in the seconds) is Write Burst End value. From that point on, the Min-Max speeds are computed for the write sustained speed range.

I'd recommend updating your shares to exclude the drive in question until you get the new one in.

Check your SMART values. Swap out the drive and let the parity rebuild it. If you're curious, run an extended SMART test after you move the files off and see if your SMART values have changed when it's complete. Alternately, run a single-cycle/pass preclear (read/write once) on the drive which should flush out any pending bad sectors.

RAID drives aren't currently supported but will be in version 3.0 though it still shouldn't hang like that. I'll take a look.

I'm pretty sure it works but I'll look into it. And you're right about the checkbox. I'll build a patch soon. I don't actually make use of the library but I know the whole thing with "Some tool you may use might use it". What antivirus/malware do you have? It looks like the input/output butter is getting messed with. I know I've recently run into issues with Acronis's protection in that it intercepts the whole buffer and won't release it until it's closed. Fucks with programs that stream HTML intermediately such as updating scanning progress. Haven't heard of IOzone. I'm using dd to read from the drive from a given location for a given duration with the bitbucket as an output.

It's supposed to preserve whatever image you put but I was able to duplicate your issue. After applying your image, you should now see a "Submit Drive" button - click on it and then click on it again to confirm. This should also allow the image to restore after a rescan.

A second text layer - should be somewhat simple to implement. I'll add it to the 3.0 To Do list. (The 2.9 line is bug fixes only now.)