Disk Errors - 3.5.1

[th3tak3n]

Current Setup:

Windows 10 Enterprise - 21H2 / Proxmox VM
Incomplete/Complete - Unraid SMB share with security set to Public
Previous SABnzbd Version - 3.5.0
Current SABnzbd Version - 3.5.1

After updating to the newest version of SABnzbd with no other configuration changes to my current setup, I have been randomly getting disk error messages for downloads. I have rebooted my Windows 10 VM, but have yet to reboot my Unraid server. Unraid is not running any major tasks, such as a parity-check. I have enabled debug for logging, though I wasn't sure the best/most appropriate way would be to post/link my log file. I tried to share a link, but with me being a new user I'm not allowed to lol.

What's odd, is I initially thought it was legitimately related to permissions on the Unraid share the Incomplete/Complete and primary library are located, but when it pauses the queue due to a disk error, I can immediately start the queue again and it proceeds without any issue until it encounters another disk error.

I'm going to reboot my Unraid server at some point today so I can rule that out as the problem, but this behavior didn't start occurring until after I upgraded from 3.5.0 to 3.5.1 the other day.

[sander]

Proxmox & SMB share ... bingo!

I would reboot, and try again with 3.5.1

If still problems: downgrade to 3.5.0 , and try again.

[th3tak3n]

Proxmox & SMB share ... bingo!

I would reboot, and try again with 3.5.1

If still problems: downgrade to 3.5.0 , and try again.

I get the complicated setup and point the finger towards it lol, but this didn't occur until after I upgraded from 3.5.0 to 3.5.1 the other day without any other changes on either my Windows 10 VM hosted on Proxmox or my Unraid share where my Incomplete/Complete folders are located. I don't mind to downgrade because I'm assuming it was something that changed in the new version, but wanted to share my configuration and issue to help resolve whatever the issue is, if not my specific setup.

[safihre]

What exact errors are you getting? Disk errors is a bit vague.

[th3tak3n]

What exact errors are you getting? Disk errors is a bit vague.

It acts like permission denied errors, but I'm able to immediately start the queue again and it continue on without any issue. I've had permission errors before which turned out to be legitimate, but this was too coincidental with the update.

I downgraded from 3.5.1 to 3.5.0 without rebooting the Windows 10 VM, and I just completed a roughly 20 item queue without any errors whatsoever. I understand my setup is probably a bit odd/complex compared to some, so it may just be an isolated issue for my specific setup. If you want/need any other information/logs/etc let me know, I don't mind to upgrade back to 3.5.1 and help troubleshoot.

It won't let me paste in the snippet of the log because it keeps yelling at me about "New users are not allowed to post links", so hopefully this image is helpful lol:

[safihre]

If you could send the whole log to me at safihre@sabnzbd.org, it's worth investigating.

[sander]

If I read correctly, your log says "Winerror 0xc000003a"

Googling that gives the below. So ... directory (disk) problem.

0xC000003A

STATUS_OBJECT_PATH_NOT_FOUND

The path to the directory specified was not found. This error is also returned on a create request if the operation requires the creation of more than one new directory level for the path specified.

[mtnbiker717]

I'm also getting the same errors with the 3.5.1 on Unraid installed on dedicated hardware, not on a VM. My Complete and Incomplete folders are on Unassigned Devices drives. My last completed file was this morning. Just upgraded the docker image a few minutes ago and I am not getting the python errors of not being able to create the folders. Here is a quick snippet of my log as well: (new user cannot post links, yada yada)

[safihre]

@mtnbiker717 your error is different, that's nothing new about that, it really looks like a permissions error on the drive.

Indeed @th3tak3n, it's a special error.. From the rest of your logs that you send me the directories are perfectly fine and already created, so why would it say it can't find the directory.. Very strange.
We did change something: we disabled buffering for faster writing speeds, but now I wonder if that's maybe causing a problem on these special systems.

We can test this theory if you are able to modify the SABnzbd source files? If so:
- Shutdown SABnzbd
- Edit "sabnzbd/assembler.py"
- Change the like: "with open(nzf.filepath, "ab", buffering=0) as fout:" to "with open(nzf.filepath, "ab") as fout:" (make sure to preserve the whitespace before the "with"!).

[Yogimax]

I have also started getting disk errors on my system, which is a Windows 10 box (not VM). The disk being written to prior to moving the file to a NAS, is a local disk with no other errors.
Again this never evidenced prior to version 3.5.1.
Cancelling the error and restarting the queue fixes the issue.

[th3tak3n]

@mtnbiker717 your error is different, that's nothing new about that, it really looks like a permissions error on the drive.

Indeed @th3tak3n, it's a special error.. From the rest of your logs that you send me the directories are perfectly fine and already created, so why would it say it can't find the directory.. Very strange.
We did change something: we disabled buffering for faster writing speeds, but now I wonder if that's maybe causing a problem on these special systems.

We can test this theory if you are able to modify the SABnzbd source files? If so:
- Shutdown SABnzbd
- Edit "sabnzbd/assembler.py"
- Change the like: "with open(nzf.filepath, "ab", buffering=0) as fout:" to "with open(nzf.filepath, "ab") as fout:" (make sure to preserve the whitespace before the "with"!).

Give me a quick and dirty on what you need.. I'm a SysAdmin, and my best "software" days are playbooks.

Edit: I did a quick search around the site, and found this link that I'm assuming I'll need to follow: https://sabnzbd.org/wiki/installation/i ... ce-windows

Edit edit: I followed the steps in the link in my previous edit, modified assembler.py and removed buffering=0, and I was successfully able to process 20 files without any issue.

[Yogimax]

I have also started getting disk errors on my system, which is a Windows 10 box (not VM). The disk being written to prior to moving the file to a NAS, is a local disk with no other errors.
Again this never evidenced prior to version 3.5.1.
Cancelling the error and restarting the queue fixes the issue.

I've reverted to 3.5.0 and had no further issues.