Older jobs with missing parts will not fail

[therealncg]

Hi all,

I've been experiencing this for a year or more and it is really frustrating. It happens on older NZB's where there is high chance of article cancellations/deletions.
sabnzb begins the download and is working through articles, speeds are in bytes per second, presumably because of deletions etc.

After 5 minutes or so it will report a few hundred megs of missing articles but will just continue on as is for an indefinite period and never fail. The longest I have left it to test is 48 hours before manually cancelling.

As a result this in progress job is stuck at the top of the queue blocking the rest of the queue and because the job does not fail Sonarr or Radarr can not then go off and search for an alternative NZB.

I've read alot and been playing with the settings in sabnzb to try to overcome the issue but it still persists.

I have 3 unlimited accounts with 3 different news providers. I have tried multiple different configurations with regards to priority and other settings but the issue still persists. Currently I have the following server settings:

news.giganews.com
priority 0
Timeout 60 seconds
Marked as required
not marked as optional

news-us.newsgroup.ninja
priority 1
timeout 60 seconds
Not marked as required
not marked as optional

news.usenetexpress.com
priority 2
timeout 60 seconds
Not marked as required
not marked as optional

Below is a short video of the behaviour. I can also provide a copy of the NZB of the job in the video.
Is there a group of settings I can enable that will force jobs like this to fail quickly? (or fail full stop!)

Thanks

https://streamable.com/9wzmdh

[sander]

At http://127.0.0.1:8080/config/switches/# ... eless_jobs , so "Abort jobs that cannot be completed - When during download it becomes clear that too much data is missing, abort the job": you should turn that ON.

Lower the number of connections to each newsserer to 30.

And why do your newsserver have different prio?

[therealncg]

Hi Sander,

Abort jobs that cannot be completed is already enabled but the issue still happens.
I will lower the number of connections of each news server to 30 to see if that helps.

I have been playing around with the priorities in order to test if one configuration is better than the other. Trying to see that if a part is not available on one server then it will try the others. Ultimately if I can overcome this issue the goal is to maximise speed which, as far as I understand, means I should have them all on the same priority but the same issue happens even when they are all set to the same priority,

Please let me know if you have any other hints.

Thanks

[safihre]

Check the content of the job, how much par2 is there? Because we only abort if the amount of bytes missing is larger than available par2.
Olders jobs used to have much more par2 available so they take longer to fail.
Additionally, the fast fail algorithm is based on the first article of each file to be missing or not. But in case of old jobs, the first article is usually present so we can't use that to decide if it's missing or not.
In this case more connections is actually better since it's quicker to try all articles.

[sander]

In this case more connections is actually better since it's quicker to try all articles.

In the videoclip:
at 0:35 the number of connections starts dropping
at 1:09 it says "timed out"

So I'm suspecting a connection problem causing a slow down.

[therealncg]

The rules say I can't post the NZB itself but I am happy to provide a link to anyone via DM if it's allowed to see if it is reproducable. It seems to happen with any nzb that looks like it has cancelled articles.

It has happened across three different ISP's. Currently on a FTTP 1000/50 connection.

[safihre]

Ah I didn't check the video.
Based on Sanders description I agree that a bad sever connection might be the problem. Your connection isn't the only thing that matters, also the distance and number of hops to your provider.

[therealncg]

Do you think that is a usenet provider issue or an ISP issue?

Ah I didn't check the video.
Based on Sanders description I agree that a bad sever connection might be the problem. Your connection isn't the only thing that matters, also the distance and number of hops to your provider.

[sander]

I will lower the number of connections of each news server to 30 to see if that helps.

And ... what is the result?

If no improvement, lower to 10.

[therealncg]

I will lower the number of connections of each news server to 30 to see if that helps.

And ... what is the result?

If no improvement, lower to 10.

Lowering connections to all servers to 30 saw no change.
Lowering connections to all servers to 10 saw no change.

[chris87]

In my instance, disabling Giganews fixed this. It appears that I was having connection issues with Giganews.

[frenchlobe]

Is there any way to fix connection problems with Giganews?

[therealncg]

Is there any way to fix connection problems with Giganews?

safihre has confirmed that the issue is fixed in SABnzbd 4.2.3 Release Candidate 2
https://www.reddit.com/r/SABnzbd/commen ... ndidate_2/