JorgeB

Don't see how.

It can't fix "unmountable: no filesystem", if can fix "unmountable: invalid partition", of course that after fixing the partition there can also be filesystem corruption. Parity might show valid but still be invalid, or partially out of sync if something was done before that might cause it, and a disk can still rebuild successfully but have some corruption, but if there's nothing you did to suspect it might no be valid it should be.

Looks like a connection/power issue, do you have another slot you could try?

Server is running out of RAM.

If parity is valid disk5 can be rebuilt, since rebuilding recreates the partition, you can test this unassigning disk5 and starting the array, check that the emulated disk mounts and contents look correct, if all OK you can rebuild on top, but if you know or suspect that parity isn't valid this won't work.

2 of the onboard SATA ports are set to IDE, that's a known issue with those chipsets, and the SSD is using one of them, change them to SATA/AHCI, there's a separate setting on the BIOS for that, usually for ports 5/6.

If was formatted it would remain XFS, formatting a disk is the only explanation I can think off that would make a full disk now showing basically empty, and unfortunately a common mistake.

That's correct, but note that the single port ConnectX-3 adapter is PCIe 3.0 x4, so it would bottleneck on a PCI gen1 slot. AFAIK no one has been able to get SMB multichannel working with Unraid yet, but it should be possible, I do have a few servers with dual port NICs, one port goes to the switch the other is for a peer-2-peer link to that server's backup, since I have no need to access the backup server @ 10GbE and this way the backups goes faster.

Does it make any difference if you set both initial and max RAM to 16GB?

Also a good idea to update the LSI's firmware since it's very old: May 29 22:40:21 TheBlackBox kernel: mpt2sas_cm0: LSISAS2008: FWVersion(10.00.08.00), ChipRevision(0x03), BiosVersion(07.19.00.00) Current one is 20.00.07.00

There are several timeout errors with various devices like this one: Jun 1 08:16:30 TheBlackBox kernel: sd 7:0:1:0: attempting task abort! scmd(0000000082b077c3) Jun 1 08:16:30 TheBlackBox kernel: sd 7:0:1:0: tag#363 CDB: opcode=0x0 00 00 00 00 00 00 Jun 1 08:16:30 TheBlackBox kernel: scsi target7:0:1: handle(0x0009), sas_address(0x4433221101000000), phy(1) Jun 1 08:16:30 TheBlackBox kernel: scsi target7:0:1: enclosure logical id(0x500605b00544d8c0), slot(2) Jun 1 08:16:30 TheBlackBox kernel: sd 7:0:1:0: task abort: SUCCESS scmd(0000000082b077c3) Jun 1 08:16:30 TheBlackBox kernel: sd 7:0:1:0: Power-on or device reset occurred ### [PREVIOUS LINE REPEATED 1 TIMES] ### Check all cables and power.

Disk5 is showing as "invalid partition", so not necessarily a filesystem problem, please post the diagnostics: Tools -> Diagnostics

There's nothing on the syslog, could be a hardware issue, but still worth to try safe mode.

Did you format the disk at any time before or after rebuilding?

That's normal it's the first bit from each byte parity is checked for every standard 4k Linux block, each block has 8 sectors (with standard 512E drives). That's seams very unlike to me if not impossible (if xfs_repair was run correctly). Yes, please do that and post the diags.

That suggests you're being limited by the device speed, not all SSDs (in fact most of them) are able to sustain nowhere near 500MB/s writes, by default Unraid uses 20% free RAM for write cache, that explains the first 8/10 seconds at max speed, then you are limited by the device actual write speed, and you can do an easy test to confirm, set the RAM cache to 40% and see if now you get approximately double the time at max speed: sysctl vm.dirty_ratio=40 Type sysctl vm.dirty_ratio=20 or reboot to go back to defaults.

There's no driver for either NIC, you can try v6.9-beta1, if still not supported it should be on the upcoming v6.9-rc1.

I would firs run an extended test on disk12, if it fails replace now, if it passes give a second change, there are also some risks doing a rebuild, e.g. if other disk fails.

Yep. You can't best bet would be ddrescue.

Mellanox ConnectX-2 and X-3 are plug play and work great with Unraid, also cheap on Ebay.

I would run a correcting check until the end, then without rebooting run a non correcting one and post diags if there are sync errors.

That's very strange, try running memtest, though I would suspect if the RAM was so bad every single sector was incorrect you'd see other issues.

SAMBA is single threaded, a single transfer will use 1 core only.

Please start another correcting check without rebooting and post new diags after it runs for a couple of minutes.

You make the backup after inserting the flash drive you your PC, you can do it manually by just coping all the files or you can use the USB tool.