Unraid OS Version 7.0.0-beta.1 available

8 minutes ago, bpwats said:

Please read this and related links in that issue. As I understand things, ZFS aggressively takes RAM and does play nice with the kernel, which only sees 5 GB allocated or so.

How much RAM do you have assigned to the ARC?

1 hour ago, JorgeB said:

How much RAM do you have assigned to the ARC?

How can I tell?

can not select zfs pool mirror mode in GUI. it's always back to raidz when I select mirror or stripe mode.. 

by the way, 8 slots

Edited Sunday at 02:07 PM by syq7970

5 hours ago, JorgeB said:

It wouldn't explain why it's trying to send an email, but it's probably because it's trying to TRIM the flash drive and failing, please post the output of:

 

fstrim -v -a

 

 

Output of fstrim -v -a:

 

 

disk3,2,1 are mechanical disks in the array (boot of course is the flash drive)

 

Note that on the nightly run I don't see it:

2 hours ago, JorgeB said:

How much RAM do you have assigned to the ARC?

Okay, I figured out a way to supply this information.

Still on 6.12.10, I created the file arcstats61210.txt with the output of command

cat /proc/spl/kstat/zfs/arcstats

arcstats61210.txt

 

 

I then installed the 7.0.0-beta.1 again, with the difference of not having ZFS Master plugin installed.
The output of the same cat command on 7 beta is in arcstats7.txt

arcstats7.txt

 

Now, the interesting thing is, the kernel oops has not taken place. The difference here, is that the ZFS Master plugin is not installed. (I was using SpaceInvader's ZFS backup/replication scheme).

 

fractal-diagnostics-20240630-1654.zipDiagnostics attached.

I saw over at the ZFS Master plugin, a really old version of mbuffer overwrites the default one in Unraid 7 beta.

 

I will try to recreate the kernel problem by installing the ZFS Master plugin again.

1 hour ago, bpwats said:

I will try to recreate the kernel problem by installing the ZFS Master plugin again.

Okay, I got the kernel hang again. One can notice easily, because the webgui becomes very slow to load (log confirms). This time with the ZFS Master installed (as well as Intel top, Intel passthrough support and GPU statistics / the latter not compatible with 7 yet.)

 

Looking at the file arcstats7kernelhang.txt I do not see something out of the ordinary RAM wise?

3 minutes ago, bpwats said:

GPU statistics

Are you saying this is not compatible?

Upgraded without a issue but now the dashboard configuration is empty?

 

59 minutes ago, SimonF said:

Are you saying this is not compatible?

I was, but actually it is. My bad.
My mistake could happen by re-installing three previous plugins in one go. GPU statistics comes first, and thereby it's requirement for Intel top is not met and I should have paid better attention to the install messages, to understand what was going on.

The default interval is 5000 msecs (on empty), with 2000 msecs being the minimum value (popover text). However, 1000 is entered from installation; this may confuse people.

 

I'm building out a couple of new systems, one destined for dockers and VM and another for strictly backups. Either of these might work well as test systems for this beta but...  It's just easier to update my main/production system.

 

So far so good after reboot, all dockers and VM came back running as expected, all disks where they should be.

 

Made a quick Raidz2 pool as a test and that seems to have worked without any fuss. I very much wish there was a way to collapse the pool's drive list in the Main tab, as I don't really care about looking at individual disks of a ZFS Pool unless i'm going to be modifying the pool or need to look more closely at specific drive(s).

There's also nothing in the Main view that indicates what kind of pool I'm looking at. Just "zfs" indicating the format of each disk, but nothing to denote it's Raidz2. You have to click on the first disk, then scroll to see "Pool Status" to learn anything about the pool.

 

Exclusive Access question:  When creating a new share, why is Exclusive access automatically set to "no" when using secondary storage, where both primary and secondary volumes each allow exclusive access by themselves (not using the "array")?

 

I'm really looking forward to finally being able to use SMB for all my shares from MacOS once I'm rid of all FUSE layers. The speed with NFS was OK, but it meant losing some attribute features.

 

Edited Sunday at 07:47 PM by Espressomatic

2 hours ago, bpwats said:

1000

Have changed this to 2000 thanks.

Upgrade for me was easy and everything started to work ok. I use ZFS pools so being able to remove the last array disk was nice.

 

A couple of people have mentioned the SSL library issue - it also affects @SpaceInvaderOne's scripts for snapshotting:-

 

Starting ZFS replication using syncoid with mode: strict-mirror
Sending incremental nvme/appdata@syncoid_Tower_2024-06-28:01:00:08-GMT01:00 ... syncoid_Tower_2024-06-30:21:13:37-GMT01:00 (~ 62.8 MB):
mbuffer: error while loading shared libraries: libcrypto.so.1.1: cannot open shared object file: No such file or directory
warning: cannot send 'nvme/appdata@syncoid_Tower_2024-06-30:21:13:37-GMT01:00': Broken pipe
cannot receive: failed to read from stream

 

Not sure if this is the right place, since technically I'm still on 6.12.10 but since this appears with this upgrade...

When you go to update OS, and then next, there is a link 'Provide Feedback on the Preleases Forums' which doesn't go to this thread, or even this forum, instead it goes here   . (Verified on mutliple servers)

Edited yesterday at 01:40 PM by Wody
altered URL

17 hours ago, syq7970 said:

can not select zfs pool mirror mode in GUI. it's always back to raidz when I select mirror or stripe mode.. 

by the way, 8 slots

It's a known issue, fixed for next release, but it's only display problem, the pool will still use what you select.

16 hours ago, thecode said:

Note that on the nightly run I don't see it:

That's normal, but it can still be trying to do it if it wasn't correctly updated, please post the output of:

cat /boot/config/plugins/dynamix/ssd-trim.cro

 

 

 

 

14 hours ago, bpwats said:

I do not see something out of the ordinary RAM wise?

ARC size looks fine, it's using the default, so you only see the issue with zfs master or other plugins installed?

13 hours ago, enJOyIT said:

Upgraded without a issue but now the dashboard configuration is empty?

Try booting in safe mode and/or resetting the dashboard config, if still the same try changing to English.

12 hours ago, Espressomatic said:

Exclusive Access question:  When creating a new share, why is Exclusive access automatically set to "no" when using secondary storage, where both primary and secondary volumes each allow exclusive access by themselves (not using the "array")?

If a share has files in two pools it can no longer be exclusive, you need to assess it over /mnt/user, or would not see the files from both pools, you can still use disk shares to access it, with the limitation that only the share files from that pool will be visible.

4 hours ago, Wody said:

instead it goes here.

That is this forum, wrong link?

19 minutes ago, JorgeB said:

That's normal, but it can still be trying to do it if it wasn't correctly updated, please post the output of:

cat /boot/config/plugins/dynamix/ssd-trim.cro

 

 

 

 

runs fine but a view error on the dashboard

32 minutes ago, thecode said:

That shows it's still using the old v6.12 way, do a dummy change on Settings - Scheduler - TRIM Settings, click apply, and then confirm it changed to this:

 

After that you should no longer see that email attempt.

 

6 hours ago, JorgeB said:

That is this forum, wrong link?

Yep, I've changed it in the post The dialog points to here.

Hi all,

 

Just a heads up. I am having issues with my AMD GPU and it's making the GUI extremely slow to load. The wait is until the AMD process is finished before allowing me to use the GUI.

 

Taking upwards of 2 minutes each time the card is accessed.

 

Jul  1 15:42:30 Tower kernel: pcieport 0000:00:01.0: AER: Uncorrectable (Non-Fatal) error message received from 0000:03:00.0
Jul  1 15:42:30 Tower kernel: pcieport 0000:03:00.0: PCIe Bus Error: severity=Uncorrectable (Non-Fatal), type=Transaction Layer, (Requester ID)
Jul  1 15:42:30 Tower kernel: pcieport 0000:03:00.0:   device [1002:1479] error status/mask=00100000/00000000
Jul  1 15:42:30 Tower kernel: pcieport 0000:03:00.0:    [20] UnsupReq               (First)
Jul  1 15:42:30 Tower kernel: pcieport 0000:03:00.0: AER:   TLP Header: 60000002 000008ff 00000061 feb02000
Jul  1 15:42:30 Tower kernel: pcieport 0000:03:00.0: AER: device recovery failed
Jul  1 15:42:30 Tower kernel: mpt3sas 0000:07:00.0: invalid VPD tag 0x00 (size 0) at offset 0; assume missing optional EEPROM
Jul  1 15:42:30 Tower kernel: pcieport 0000:00:01.0: AER: Uncorrectable (Non-Fatal) error message received from 0000:03:00.0
Jul  1 15:42:31 Tower kernel: pcieport 0000:03:00.0: broken device, retraining non-functional downstream link at 2.5GT/s
Jul  1 15:42:32 Tower kernel: pcieport 0000:03:00.0: retraining failed
Jul  1 15:42:33 Tower kernel: pcieport 0000:03:00.0: broken device, retraining non-functional downstream link at 2.5GT/s
Jul  1 15:42:34 Tower kernel: pcieport 0000:03:00.0: retraining failed
Jul  1 15:42:34 Tower kernel: amdgpu 0000:04:00.0: not ready 1023ms after resume; waiting
Jul  1 15:42:35 Tower kernel: amdgpu 0000:04:00.0: not ready 2047ms after resume; waiting
Jul  1 15:42:38 Tower kernel: amdgpu 0000:04:00.0: not ready 4095ms after resume; waiting
Jul  1 15:42:42 Tower kernel: amdgpu 0000:04:00.0: not ready 8191ms after resume; waiting
Jul  1 15:42:51 Tower kernel: amdgpu 0000:04:00.0: not ready 16383ms after resume; waiting

 

Log is full with these messages and it is a pain to get into the VM tab, VM settings, Tools->System Devices.

 

I had issues with this card on v6.12 but it never interrupted the browsing experience.

7 hours ago, JorgeB said:

If a share has files in two pools it can no longer be exclusive,

Yes of course! That makes sense, as I'd forgotten that both pools together constitute the share and you have to be able to see the merged filesystems or you'd possibly be missing files at any given time.