TL:DR: I think I've got it fixed. But, wanted to answer questions in case others search and need help with similar issues.
evilhero wrote: ↑Wed Jan 27, 2021 3:27 pm
So within the directory that it's trying to delete, what are the permissions / ownership like therein? Is there just the cbz file or are there additional files in there as well (I'm talking about a individual download location, not the root of the SABnzbd completed download location btw).
Everything's 777; open to everyone for anything. It was just the CBZ file. Mylar would see it, move it, and when it tried to delete the parent folder, error. So, it should've been pretty simple I would think.
evilhero wrote: ↑Wed Jan 27, 2021 3:27 pm
So because Mylar is in docker and SABnzbd is in docker, it's considered to be on different machines - this is more from a general standpoint that one specific to your exact install. We usually say tell individuals running dockers to make sure that the ownership/perms are fully open to both users running Mylar and SABnzbd to the SABnzbd completed download folder and to point the Completed Download Location for SABnzbd (within the Mylar program) to the accessible folder location from within the given docker.
I thought about this too. I even tried to check the box for SAB on a different machine and that actually made it worse. Once I did that, it wouldn't even find the file much less attempt to move it. Errors abound. Perhaps if I was running them as independent containers, I could see having to specify they're on different machines (different networks, file systems, etc.). In a stack, they should see each other easily (and they do for the most part).
evilhero wrote: ↑Wed Jan 27, 2021 3:27 pm
The error in question is being thrown because the shutil.rmtree (which is a bundled python module) is failing for some reason, none of which seems to be logged unfortunately - so we need to figure out why the function is failing to remove (which could be for any reason really).
I saw this in some of the code when I was troubleshooting. Perhaps adding some additional logging for such a case? I saw in the code that this particular piece of code didn't have it's own try/catch even though most everything else in there did. The error that I see in the logs seems to be a catch-all at the end of all the code.
-----------------------------------
So... Somehow, I fixed it. And I'm not really sure how.
I mentioned before that I was using the Docker image from Hot.io. I thought that I might try the image from linuxserver instead. So, I updated my Docker Compose file to use that image. Unfortunately, it wiped my config when it updated the container. Probably due to rebuilding of the bindings? I'm not sure. Either way, I had to go through and setup everything again. I tried to exactly replicate what I had before. Once I got it all setup, I tried that comic again. And it worked!
Moved the comic
and deleted the parent folder. I tried a few more comics just to make sure and it's all copacetic now.
So, what was it? I have no idea. Perhaps the new image? The SAB image I was using was also from linuxserver. Maybe that's it? As far as I can tell, everything is the same as it was with the Hot.io image: permissions, settings, folders, etc. Maybe it's just better not to mix images from different repos.
I'm good now. Wanted to at least report on what I went through in case someone else runs into this too. Thanks so much for the speedy replies and trying to help out. It is much appreciated. And a great program.
Cheers...
EDIT: One other thing I wanted to mention that I noticed... When a comic fails to download in SAB, it seems to stick around even though Mylar sees it failed and tries again later. Sonarr/Radarr detect these and generally wipe them from SAB so that SAB's log doesn't get full of failed downloads. Is that something that Mylar should be doing that I'm not seeing? Or is this a feature request?