'Off The Rails' Event Being Monitored, Appears Some Peering Is Creating 'Skips'

[Beck38]

I monitor three major newsgroup providers (Astraweb, Giganews, and Blocknews), both Us and Eu servers, and although things have been going very well the past 2-3 months with no major 'skip' (or 'fade') events, starting about some 12 hours ago I noticed things were getting... well, a bit off.

Right now, Astraweb/EU is skipping tons of parts posted in the last 24 hours, as is Blocknews/US and /EU, but neither Astraweb/US, or either of the Giganews (US or EU) have suffered any (as far as I can tell). Somewhat obviously, it appears that however (or whomever) Blocknews and Astraweb/EU are peering from, is giving them the problems.

If it continues for another 12 hours or so, and doesn't appear to be 'filling in' the skipped of missing parts, I'll do some more research and try and nail which provider is causing the skipping/Fading.

Sometimes the peering, if done properly at the server, will 'fix' this, but this is the first time in probably 6 months that I've seen multiple server plants be affected all at once, while others don't seem to be at all.

Just a heads up, if you try d/l'ing anything posted during this time-frame, you may have serious problems with missing parts that may or may not be 'fixable' with the pars as the amount of missing blocks may well exceed the available pars.

[Tokeman]

Hopefully they sort it out soon. I'm on astraweb, and in the us, but when I connect to the SSL server (ssl-us.astraweb.com) it always ends up connecting to a EU server for some reason... Also using blocknews for my backup :\
Thanks for the info. Keep us posted!

[Hypatia]

Astraweb has that sort of problems on EU server from time to time.. US works fine in this case

PS it seems it got fixed...

[Tokeman]

Anyone know why my US SSL server address (ssl-us.astraweb.com) always ends up going to EU when it connects?
Thinking about mailing support, but figured some one here may know.

[Hypatia]

tobh i have no idea. ive just made tracert and i have different routes to EU\US servers, and different IP addresses. What does yours show?
did you try to use servers IP address instead of "ssl-us.astraweb.com" ?
there are quite a few..

[Beck38]

AT just around 24hours since the initial discovery of problems, most (but not all) of the monitored servers have recovered. A bit of what I have discovered:

Giganews/Us and EU, plus Astraweb/US, suffered zero problems.

Blocknews/US had gaps about as severe as anyone, but recovered fairly quickly.

Both Blocknews/EU and Astraweb/EU still have major gaps that have not filled in as of yet at the 24hour plus point.

All servers that suffered gaps were utilizing Newshosting as their primary or secondary peer. What was interesting was the fairly rapid recovery of Blocknews/US, and it appears that they utilize several primary peers, that enables them to fairly rapidly recover from any peer that 'goes off the tracks' extensively. Exactly why their EU server doesn't do the same... They may be contracting out that portion of their system, and the operators of that haven't programed it up to the same level as their US server.

Similar things may be going on with Astraweb/EU. I'll continue to monitor over the next couple of days.

Once again, this shows that if one is to put a secondary server into ones binary d/l'er, the secondary selection is to put the US or EU server (opposite of your primary) as your first choice (costs nothing of course). The operators SHOULD have those two 'locked' together, where each is the primary peer of the other; but obviously that is not the case.

Except with Giganews; I can think of some drawbacks to peering each plant to the other, and in the past (in fact, the last major 'event' 6 months ago was Giganews related) and the gap then never (monitored for 2+ weeks) recovered, but what was most interesting was how fast Blocknews/US seemed to recover from the 'hole'.

I'll see if those two servers lagging in filling up ever manage to recover in the next couple of days.

[Hypatia]

to be honest i had the impresison that astrawebs' "main" server is EU(both general news.astraweb.com +eu.news.astraweb.com are in Europe
). And i think i wont be mistaken if say that majority of people consider AW mostly a european provider.
That is why i find the problems they have from time to time on EU \general server kinda surprising..

[Beck38]

Yep as to Astraweb/EU supposedly being the 'main'/primary; maybe they simply hired a crew to do the San Jose plant are 'on the ball' more.

Since I used to live right down the road (okay, a a few miles) from where they're physically located, one would think the technical ability of folks walking down the street are 'above normal'. But who knows. Again, I'll continue to monitor for a few days as usual.

[mesaman]

Anyone know why my US SSL server address (ssl-us.astraweb.com) always ends up going to EU when it connects?
Thinking about mailing support, but figured some one here may know.

I think you're looking at a list of message-ID's for the post you're downloading. It's a common misconception, if you're using a newsreader that shows the message-ID's you're downloading, to think you're connected to the server that appears in the message-ID's.

[Tokeman]

I think you're right mesaman.

[Beck38]

72 hrs later, and both Astraweb/EU and Blocknews/EU have yet to 'fill in' the gap at all.

The 'event' lasted approx. 12 hours, with about 35-50% of the blocks being skipped.

Will continue to monitor for a few more days, but after 3 days, I doubt that any missed parts will be 'filled in' like Blocknews/US was.

[zot]

Beck - Are the headers or bodies missing? Or both? (I've found that the headers on older posts - especially in text groups - are much more likely to go missing than the bodies )

How does a free server such as Just4Today compare?

I've noticed that Highwinds, Readnews, and Astraweb all appear to use the other US/EU server for backfilling missing articles on old posts, but I don't know how long the process takes. I think that Astraweb only had an American server until about 6 years ago when the Amsterdam location was set up. For several years, Highwinds had consistently better completion on the EU server, but now completion is identical - at least on old posts. (If a 500-day-old file has 73 parts missing from Highwinds' US server, the EU server will be missing those identical 73 parts) So Highwinds is obviously back-filling between server locations and therefore it would be pointless to setup Highwinds US and EU servers as primary/secondary anymore.

It's interesting that the US servers seem to be in better shape now than those in Amsterdam, since for many years it seemed the opposite was true, at least in my experience. (Giganews being the exception)

[Beck38]

Yet another large 'gap' today, starting around 2000hrs GMT or thereabouts. Astraweb/US, Giganews/US no problems, and with some gaps but fairly quickly filled (within an hour) at Blocknews/US. All three EU servers of each, the gap is still going on (not filling in).

All the original gaps from 96hrs ago, no fills whatsoever on Astra/Block/EU. I'm willing to bet they won't fill in at all.

Today's gap lasted at least 4 hours, I'll take a hard look in about another 4 to see if anything changes.

FYI it's been years and years since I looked at any 'free' servers, sorry.

I might add that except for extremely minor, and usually filled in within an hour or so, these are the most major 'gaps' I've seen in several weeks to 3-5+ months. Are their personnel starting to take spring vacation, or what...?

Fast Update: it appears to me, that Astraweb/EU has not been updating for well over 6+ hours now (I try to 'update' the article scan, and get 'no update since last scan'). No fills on any other EU servers, including Bocknews/EU and Giganews/EU. Something out there has completely broken down. All US servers appear to be and continue to update.

[Hypatia]

did you write to support explaining the situation?

PS When i download headers the speed is awful

[Beck38]

Minor 'skips' still occurring in 'near real time'.

The three parties mentioned have been told, but I'm sure that I'll get the same response I have in the past, that usenet is not a 'perfect system' (of course not, but...), and that's why they always recommend a secondary server.

Why the companies don't at least do 'database rectification', in making sure that their US and EU servers are more 'in sync', is beyond me. That would eliminate a lot of the gaps I've seen over the years, whether it's US or EU with the gaps, and the other non-gap'ed. Although it's very interesting that Bocknews/US seems to both have more gaps yet 'cures' them very quickly. I'm seeing the same thing occurring with Giganews/EU, where it takes some 6 or so hours, but it 'fills in'.

Astraweb/EU is still 'lagging' 12+ hours behind Astraweb/US. Hasn't filled in any of the missing headers but I'm talking 'real time' where the Us server is v. the EU one. Weird.

[newcster68]

I really don't know what you all are talking about with these so called gaps but is this the reason why some of my shows that i am downloading are not complete? Checking files and all are there but when downloading them it is only downloading like 375kb per file...wtf? what is going on? Im on Newsgroup Direct

[Hypatia]

some of my shows
Newsgroup Direct


it happens most likely due to DMCA takedowns. Newsgroup direct uses highwinds servers as far as i know
several shows are taken down extremely often on giganews and highwinds

[zot]

I really don't know what you all are talking about with these so called gaps but is this the reason why some of my shows that i am downloading are not complete? Checking files and all are there but when downloading them it is only downloading like 375kb per file...wtf? what is going on? Im on Newsgroup Direct

375 KiloByte is a typical article size, so I assume that you have one (complete) article per (rar) file - kind of like what HBO does to submit DMCA takedowns :P

You can also look inside the NZB file (open it using any text reader) to check if it looks like all the articles are there.

Binaries are posted as articles, and these articles are supposed to be automatically grouped into rar files, and these rars in turn are automatically grouped into releases. But sometimes the system breaks down and auto-grouping does not occur.

[Hypatia]

Astraweb replied to me that they are doing "maintenance on the headers" =)

[Beck38]

That's the same 'announcement' I got.

Vis-a-Vis Astraweb/EU, the Major gaps from the 13th and the 17th are still there, haven't 'filled in' at all. Some minor wobbles from the 18th, 19th, 20th, have partially filled in, the same with Blocknews/EU. But the only servers without the 'gaps' from the 13th/17th are Giganews/US and EU, Astraweb/US, amd Blocknews/US (the latter after filling in over some 2 days).

I doubt very highly at this point that, in particular, Astraweb/EU will fill in at all. I'll try to watch for the next week or so, but the gap(s) are very extensive, near 100% loss of parts for hours at a time.

[Hypatia]

did you inform them about the gaps? And that US server is Ok? i cant believe these f ...rs wont do anything even having all articles on their other server available..grrr

[Beck38]

Well, believe it or not, as of around 0700GMT today the two 'laggards' (Astraweb/EU and Blocknews/EU) finally caught up completely...

I almost fell off my chair. It took them quite a while, far slower than Giganews or even Blocknews/US, but they DID catch up; only took close to five days.

If anything, it shows (once again) not to 'jump' on things right away, wait a bit for the propagation and the server 'fill-ups' to occur.

[Hypatia]

it shows (once again) not to 'jump' on things right away, wait a bit for the propagation and the server 'fill-ups' to occur.
and\or set AW US server as a fill one ;)


PS Headers on AW are way behind actual articles available on the same servers.

this "maintenance" is taking ages...