many downloads incomplete - astranews

[Beck38]

If so (and that quote isn't to be found on the Astraweb site, btw), I'm monitoring the three top usenet groups (top as in the amount of traffic in GB as listed by several tracking sites), plus a smattering of others that were listed at the top of this thread when things were first noticed 'going off the tracks'.

The only notice put up on the site so far as to any problems, was exactly one month ago. No updates, no nothing, since. So I opened another ticket asking 'what's up?'.

Like I said, it's been a month since the last 'official' pronouncement. The skips, according to folks here and elsewhere, starting showing up probably two weeks before that. The fact that both server plants, US and EU, were affected pretty much exactly the same, is to me the biggest clue as to what happened, either one plant ran off the rails and then propagated the failure to the other (an original h/w failure made worse by a s/w that failed to do proper checking before propagating it to the other server), or a upstream peer that nailed both servers simultaneously.

But I'm sure by the end of the days, folks simply want it fixed. From what I see, not anytime soon, and pretty much becoming never. Sad.

Got a reply to my message on their site. 'Will be fixed real soon now'. ha ha

[Beck38]

As of some 5 hours ago (from the date/time of this message) still nothing changed.

Folks are getting antsy (uh, me?), and there are several folks/threads elsewhere, where folks are reporting problems where there doesn't seem to be any (mine and other confirmations of such), where users are making simple mistakes and then blaming it on Astraweb.

Again, they better fix things soon, like right now, or it'll snowball.

[cola]

Do you have an NZB that you know won't work on Astraweb? I'd like to test around a bit.

[Beck38]

Lots been listed here, just look at the first few posts.

The last 'news' posting by them was on 26June, saying that the gap was centered around 110-130 days, which would have put it around the end of February. Since that notice was just shy of 30 days ago, the gap as moved (time marches on), and center should be around 160 days back now.

I use the 'mergekit' HD postings as a way to see what's going on, since I d/l'ed them all (through Astraweb/US) at the time them were posted; therefore I know they were 100% at that time. Generally the 'mergekit.C' posts are the smallest, and any skipping is a dead giveaway as to problems with their plant.

However, tonight I notice that Astraweb/US is not adding any files, either ones that are posted at them (they are being propagated out to other servers, through) or ones coming from their peers. That tells me that 'maybe' they may be working on the plant (finally). FYI Astraweb/EU is not being affected. That tells me the 'front end' of the plant is still up and running, but the 'back end' may be shut down for 'maintenance'. One can hope, we'll see by tomorrow some time.

[cola]

I was hoping something for something more specific. I tested a handful around 160 day mark and couldn't find anything. I did find 2 that were had incompletes around the 70 day mark. EU server didn't seem effected. Interesting part is that using different ports gets different incompletes. Setting the EU server as the fill server seems to almost get around the problems on the US farm.

[PastTense]

" or a upstream peer that nailed both servers simultaneously. "

I don't understand this at all. The theory is that you have multiple peers. Each post is going to be offered to you by multiple peers--you take it from the one which offers it to you first. So if there is an upstream peer which isn't offering posts, you just get it from someone else. And a major site like Astraweb should be very well connected.

So could someone explain this?

[Beck38]

Depends on how your s/w is set up. It used to be you defined a 'trusted' peer, and got 99% of your stuff from there, and only any others if it was something 'new' and unseen.

Still no recovery as of today, I'm seeing more and more folks dumping them and going with whoever.

[zot]

I don't understand this at all. .... So could someone explain this?

Just speculation.

I can explain what I think the problem might be:

The major companies like Highwinds and Astraweb appear to use some kind of automated software that back-fills missing articles between the US and EU servers. I've noticed that in many cases the US and EU server will be missing the exact same articles on older files -- files that are 100% complete at other companies' backends. Steve at Astraweb has stated in the past that they back-fill missing articles across the Atlantic between server farms, but I've never seen him post the details of how this is done. (trade secret perhaps?)

But anyway, it must be some sort of automated process, because there are far too many articles on a server to cross-check/back-fill manually (and let's assume that they don't have a huge staff).

Because both of Astraweb's server farms were suddenly afflicted in the exact same 100-day post range, I strongly suspect that the problem was a result of a malfunction of this automated cross-checking/back-filling software that Astraweb presumably uses.

Highwinds appears to be set up in a similar way: all older (500+ day) files are incomplete, and the US and UE servers both have (mostly) identical missing segments.

This is why it's so important to set up a block account (that uses another backend) as a fill-server. In the old days (3+ years ago) it seemed that server farms were completely independent, and using a provider's EU server as a backup to its US server would usually deliver 100% completion on old (non-DMCA'd) files. But today, that's no longer true: whatever is missing on one server will typically be missing on the other.

[cola]

[QUOTE=zot;3597434]

In the old days (3+ years ago) it seemed that server farms were completely independent, and using a provider's EU server as a backup to its US server would usually deliver 100% completion on old (non-DMCA'd) files. But today, that's no longer true: whatever is missing on one server will typically be missing on the other.

You'd be surprised, looking at one file around 70 days ago, I got these results using the NZB completion checker:

This was the US server on port 119


Port 23


Port 1818


Port 8080




EU Server, only port 119 had an incomplete

[zot]

Thanks, Cola. I'm completely stumped on why the port number can make any difference at all -- assuming that your result is consistent and repeatable across multiple trial runs.

Since we know the same backend is used regardless of the port setting, that would of course suggest that the article exists on the server, but is somehow not being reported correctly.

Have you tried checking the completion using SuperNZB? It uses a different method, which appears to be more accurate, faster, and overall superior in my opinion.

https://filesharingtalk.com/threads/4...nt-mini-review