many downloads incomplete - astranews

[Beck38]

Not that I can see. The 'barrier' at the top (most recent) end is around 83 Days (9 Apr 11), things newer than that are okay (as far as I can tell).

From 83 days to around 300 (2 Sept 10), pretty much hosed. The missing parts do die off in a slow curve, going from the really bad in this April to just a bit towards that 300 day mark. Since I just completed that testing, I don't know if the plant is 'improving' at all, will have to wait and re-do things a bit in the days to come.

Of all that I can accurately test beyond the 300 days looks okay, but like I said there is a 'curve' of missing parts from this April going back, and by the time one gets to around 200 or so it's pretty minimal, IF the poster put up a decent amount of pars. But, I tested my own postings, which I know were as 100% as possible (JBinUp and header tests on multiple servers both Astra and Giga) when posted. Like I said, if it was worse I don't know, we'll see in a few days if things start improving, and from which 'end' (newest or oldest).

[zot]

I never came across this problem at all. Maybe because I had been using Binload for everything 400 days and less (and getting 100% completion) and saving my AW block for the older posts - which only had an occasional incomplete.

The last time users were reporting that Astraweb had a serious completion problem was around Christmas 2009. (I was too busy to investigate it then, and now this one pops up at another major (US) holiday) Judging from the last meltdown and its eventual resolution, I'm guessing that it might take AW a few weeks to back-fill all the missing parts, we shall see.

[Beck38]

That last 'major' meltdown I remember well, and to my mind they really didn't recover completely from that; I believe it was related to some plant upgrades that went awry. Others have also commentated elsewhere that they see skips and such on 'known good at one time' postings going back 250+ days, so one can hope that they don't simply fix the 'minimum' range and leave it at that, which has been the result in the past.

Again, I find it of interest that both servers would go off the tracks in virtually identical ways; one hopes that that is at the top of their priority as to investigation.

In doing some re-tests today, no difference in number/severity of skips, in the 'hit zone' of 100 days; the 'roll-off' seems to have changed, but it's within the 'margin of error', no real change.

Doesn't look like any work is being done.

[Beck38]

4 JUL Report - No changes, no fills, no corrections noted

[Hypatia]

you should send them email.They should be bombarded with emails of the same type: FIX UR SHITE ASAP FFS!
of course all the necessary data to back up this statement(you have problems, guys) should be there as well

[Beck38]

They have been, and have a 'note' (wasn't this previously listed?) up saying they are 'working on it'.

http://helpdesk.astraweb.com/index.p...news&newsid=53

"Incompletes & Article Not Available
Posted By: Steve On: 26 Jun 2011 5:17 PM
Details Incompletes & Article Not Available

We are looking into and fixing the issue with article completion for articles at and around the 110-130 day old mark. This issue is unrelated to the Network Instability issue posted earlier today.
Completion will return gradually as the articles are recovered.
We regret and apologise for this incident, and our engineers are working with urgency to rectify the problem."

I'll continue to check known original 'completes' at that 100 day mark, ones further back around 250 days, and those around 75 days ('gap creep') and for both any additional gaps or any 're-fills', if any. So far, none noted.

As both plants have pretty much the exact same gaps, they can't use one to 'cure' the other, and that being said, the core of the problems (I think) is one of their peers. Identifying which one, then fixing the actual data, is going to be a big project.
Unless this happened really fast, all at once, means that their watchfulness was a bit too minimal (I ran a usenet server plant for close to three years in the late 90's).

Good luck.

[Malcontent]

http://forums.newsbin.com/viewtopic....=28479#p175011

by stevef
Official Astraweb Representative

The wheels are in motion. We had some recovery early last week, then a few days of non-recovery later last week as replacement hardware was ordered. The long weekend doesn't help.
We expect the hardware and other things to be in place in a few days, and the recovery will continue then. I'm not intimate with the recovery process, but I'll post more if and when I get more details.

The post it near the bottom of the page.

[Beck38]

That's interesting that they are going down the h/w failure route, but most of the plants these days are built using some pretty specialized h/w. I did have some (somewhere) some interesting pictures of some SATA Rack-Bays and such...

So, I guess I'll wait until the end of the week (TH/Fri) before running any more tests.

[loubarlow]

Has anyone figured out a formula to discovering what age of articles are affected? Is it between a certain timescale or *all* articles of a certain age?

[sandman_1]

by stevef

We are well aware of the problems and we are fixing it. The estimate for the window of incompletes may have been a little off, but it was an estimate. What is important is that we are addressing the incompletes, regardless of whether it is 1 day or 100 day window.
Although the barrage of complaints is well deserved, it does not help to speed up the recovery.

I would disagree with stevef. IDT he has heard of the term, "The squeaky wheel gets the grease." I think if anything, the more complaints, the faster they are going to fix it. If they start pissing off tons of their customers where they might jump ship, you bet that will light a fire under their ass to fix the problem quickly. And I think Quade needs to sthu sometimes. If they are advertising a certain percent completion and people aren't getting it, then they have every right to complain.