an alchemist?
Juice boxes. Are you getting lots of juice boxes? While you're out, can you pick up a roll of aluminum foil? Oh, and some plastic bags. Wait, no, I'm picky about bags. Fuck it, just the aluminum foil. You know what, I could use some juice.Originally Posted by Beck38
Everything is brought to you by Fjohürs Lykkewe.
At over 25 days, I don't see any improvement whatsoever in any of either the 'test' file posted or on any previous files posted in the last 6+ months. Astra sent me an update on my ticket saying that they've 'fixed' everything. I don't see it at all.
Just for 'fun' I re-posted the exact same test file (but in a different newsgroup) but through Giganews. We'll see how that fares on either US or EU Astra.
UPDATE: Came through 100% on Astra/US and EU (So Giganews propagated and Astraweb received just fine). Now we'll see (once again) how the files degrade over time (most probably).
Last edited by Beck38; 08-27-2015 at 05:28 AM.
No change after 24 hours on the new posting (still 100% on both US and EU).
File posted 30 days ago still pathetic yielding around 30% corruption on both US and EU.
I gave the Astraweb folks a fair piece of advice on how to track server corruption, maybe they'll take it or maybe not.
I get better completion rate these last days.
Which nzb, how old, which server? Virtually nothing I d/l from about 2+ weeks old to 6+ months is pretty hosed, same from either Giganews or Blocknews are 100% complete.
elementary s03 DVDRip from a couple of days ago was 100% complete. In astra EU.Which nzb, how old, which server? Virtually nothing I d/l from about 2+ weeks old to 6+ months is pretty hosed, same from either Giganews or Blocknews are 100% complete.
You just answered the question; if the file was posted only 'a couple of days ago' then of course it's 'ok'. Keep that nzb and give it a try 2-3 weeks from now, you'll see distinct errors creeping in. Give it a month or two, and unless it has a huge amount of pars, unrecoverable.
UPDATE:
The Giganews posted test tiles took their first errors, right at 5 days in. .1% (both US and EU), but there nonetheless.
If it follows the previous tests, it'll start getting really bad (>10% errors) around 10 days, and drop off a cliff around 20 days (>20% errors).
Those earlier test files are getting really bad, exceeding 45% errors at the 30 day level.
NO response from Astra as to my latest update (4 days ago) to the ticket I originally opened.
Last edited by Beck38; 09-01-2015 at 12:46 AM.
I understand now what the problem is. I will try those nzbs in a few weeks.
Exactly how the retention degrades is, to me, a really good question. It's been 20+ years since I ran a usenet plant (either did so or provided technical assistance to folks going back to the mid-80's), and it appears that the Highwinds group got tired of whatever they were using several years ago, and spent the money to develop their own NNTP software, which I gather they transitioned all of those plants they bought up to that 'standard'.
Giganews in particular is proud of their own 'in-house' software development team; I don't think they have it available on the open market for any other provider.
Don't know what Astra uses, there's no information as to what they run (on their website), although there are copious 'upgrade' messages abounding ever few months going back years, so they may have either some software they developed themselves, took an older package (that's no longer actively supported), and are keeping it running by making 'improvements' (my guess), and what is causing these degradations (that keep happening at pretty regular intervals) may be something buried somewhere in the original code they haven't yet figured out.
If so, they need to have a very rigorous error checking program to catch when things begin to go 'off the rails', and quite obviously they don't. What's interesting to me is that as things go south after a few days/weeks, it stays pretty bad for several months, then going back further things 'magically' get better, at least the things posted 6+ months ago do.
It may be that their plant s/w transitions posted articles from 'near-term' storage to 'long-term' after a few days, then to even 'longer-term' after a month or more, and that's where the errors creep into the 'system'.
Whether or not things posted right now will get better (after getting very much worse), is probably a crap-shoot, and probably (!) they'll figure out what's causing the degradation before we hit that point. I have, upon occasion, done reviews of 'cloud' systems that were being run for large corporations, and found that the checking programs they had in place had holes one could drive a Mac truck through, and from what I see with Astra, it's about the same.
Last edited by Beck38; 09-02-2015 at 12:20 AM.
I've been reading your reports with interest Beck38. I'm sure you don't need any 'atta boys' but I do appreciate what you are doing and the time you have spent monitoring astraweb. Thanks for posting your findings.
Posting Permissions
Reply With Quote
Bookmarks