SABnzbd Forums

Can we get a new optional switch to "Abort jobs that cannot be completed" that isn't based on percentage of missing articles? My queue stacks up so quickly, that reasonable data capturing in large quantities is not possible; jobs that are very large (20GB+) suck up too much time and data in my queue before they get aborted. I notice that most downloads that will complete will Ramp Up in Bandwidth very quickly at the start (5sec); and those those will ultimately fail will Ramp Down in Bandwidth and begin showing missing articles. Thanks.

What version of Sabnzbd are you using?
Because it hasn't been based on a percentage for quite a while. It's based on the first articles of each file that are tried at the very start. It's called fast_fail in Config Specials.

However, ifn you have many servers, all methods take long because we have to try every single server.

Using 3.1.1, with Fast_Fail ON, Abort if cannot be completed ON and Req_Completion on DEFAULT.

This 2600MB download below took roughly 3min to nearly complete and then fail.




Multiply that timeframe for larger files (26GB => 30min) and they can tie up the queue for long periods with back to back failures.

Fast_Fail appears to work great for those nzbs with poor first articles, but I'm assuming these scenarios have issues later in the process if the nzb is not being aborted immeadiately.

I'm currently having to stalk my queue and remove the failing downloads; while sending the good ones to the bottom. I am averaging 80MB/s for throughput, so every minute I'm not on a good nzb I'm losing roughly 5GBs of data. The pattern I am noticing with the nzbs that are "hanging" is that they bandwidth drops below 1MB/s, would it be possible to signal an ABORT anytime the bandwidth drops below 1B/s for 3sec continuously? Or a setting that lets users choose the threshold (Bandwidth or Number of articles) in which they want to abort; such as any download below 5MB/s would be considered a fail regardless of missing articles? Thanks.

So far I've never seen cases where the first articles are fine, but then the rest suddenly isn't. I'd like to investigate what is going wrong.
Could you share one of the NZBs? You always see this happen exactly like this?
Another reason this could happen is if a job is retried, but I guess that's not the case here.

How would you like me to share the nzb? Thanks.

You can send it to me at [email protected]

Any insights to my issues? Here's a screenshot of my queue after it encounters a number of consecutive "hanging" nzbs; note the severe droughts of data in between attempts.

Just to be sure, can you enable Info logging in the Status and Interface settings window and then download the log after a while?
Can you check what it shows? The long breaks seem odd.
Due to regular work days I haven't had time to analyse the NZBs yet.

This is a fairly common pattern when some articles are missing and some are there. The server actually takes a lot longer to respond for each article when it's missing. You could try using the servers from a specific country instead of the default news.usenetserver.com. Some of the Highwinds server farms respond a lot faster than others. If you can increase the number of connections it will also go faster.

I was using the default UsenetServer @ 20 connections; tried the US-UsenetServer @ 20 connections with no change. I am only able to download approximately 1TB per day on average; I should be able to get closer to 4TB. Is it possible to add a "Missing Article Limit" to the switches page? It generally takes 5 seconds to determine visually if a nzb will complete or fail.

Have you tried lowering the req_completion_rate to something like 90%?
ALthough this indeed does not fix the multi-episode downloads.

I made a suggestion for a fix here: https://github.com/sabnzbd/sabnzbd/pull/1785

If you would like to try it you can download it here: https://github.com/sabnzbd/sabnzbd/acti ... /540666965

Go to Config/Special and set missing_threshold_mbytes to for instance 200. After having tried to download 200 MB it will fail if more than half is missing.

Puzzled-

Can you walk me through the install process? It appears that you edited a couple of different configuration files.

Sorry, it seems you'll have to be logged in to Github to download them. I assumed they would work because the page is visible for anyone. If you are logged in you can click on one of the links listed under Artifacts to download an installer for Windows or the Mac binary. It's only available as source for other platforms.