Wierd Errror, 'TooSmall' Occuring On Astraweb/US-SSL

[Beck38]

I have a ticket in on this, but in well over 20+ years of being on Usenet (on more servers than I can count), I've never seen this type of error:

441 437 Rejected 3 000001 TooSmall

Like everyone else (I'll 'assume'), I get my share of one connection or another (I usually run 8 SSL connections) were the 'front end' processor gets either 'confused' or whatever ('Socket Connection Reset'), usually get those on average once an hour (although sometimes it runs for several hours with no 'reboots' on a connection.

But this is the first time I've ever seen a 'TooSmall' message. Still running JBinUp 0.90 Beta 8 Build 738, going on what, over two years now.

Like I said, got a ticket in at Astra asking 'what kind of error is this?' but... had Mr. Google look for a reference, and the only real return was way back in 2006, where folks are discussing it as happening 'all the time' (again, I've never seen it in 20+ years!)

Who knows. These arn't 'small' files by any means.


UPDATE:

About 10 minutes after I posted this (and my query to Astraweb) my ISP went down hard. HARD.

It's possible that they had something going on in their plant, it's been 'shaky' the last couple weeks (at least 2 'scheduled' repair outages in the middle of the night that didn't seem to affect me).

We shall see.

[Hole69]

1) Dump J-bin (beta!?) and try Sabnzb+

2) Run Memtest and a HDD diagnostic to rule out local hardware issues

3) Bitch to your ISP, it looks like its their fault.

4) Split 10/10 connections between Astra US and EU. This seems to be the most stable config.

If you have Scocket Connection Refused something is wrong somewhere, you should receive no errors constantly at all.

[Beck38]

Jbinup is an uploading program (the only one I'm aware of that does error checking with the usenet server); Sabnzb is download program.

After things came up, it seemed to stabilize for about 30 minute, then I started getting the errors again. A quick call to the ISP and they still had techs 'out in the field working on it', so I let it run. 'Noise' in the cable plant, so they said.

But no errors of any type for the last 2+ hours, so it may be they found (and fixed) the actual problem, but time will tell.

[mjmacky]

I haven't uploaded in a good long time, but maybe I'll run a test myself to see what's up.

[Beck38]

I'm about 99+% convinced it's my ISP (Cable/Comcast), as they yanked the system (local neiborhood), got it back up really quckly, the errors continued for a while, then faded out.

But it comes and goes. I live right next to an Airport (county/industiral, largest airplane manufacturing plant in the world) and tons of potential for 'ingress' from VHF-AM and high-power HF (aircraft frequencies) which are well known to cause havoc with cabletv frequencies. But except for a couple of outages over the past two years, this has been the only one I could point to very specifically as causing some kind of 'error'.

I gave the Comcast folks my 2cents, hopefully they'll spend some more time tracking down the problem, as it continues intermittantly every 2-3 hours or thereabouts, since they 'thought' they had 'fixed' it yesterday.

I just noticed something in my logs that's VERY interesting.

The error rate (and therefore the 'toosmall' error) jumps way up almost exactly every three hours - like clockwork.

I tracked it from this afternoon to this evening, and yep, every three hours ***ZZZZZAP***

Then it settles back down again, and runs smooth as silk until the next three hour point comes up, and.... you guessed it.

It's really beginning to look like a timing problem with the network, where something is slowly getting further and further out of sync, until finally it 'losses it' (large amount of errors ensue), and then it 'snaps back' to being close in sync again. Then slowly drifts out, with little if any errors, until it once again gets far enough out that it can't quite recover the data stream and....

Plop. I really wasn't watching it, or the logs, that closely before, and I'm sure that virtually nobody else is really bothered by it at all (approx. 10 minutes of variable throughput every three hours). But there it is. I passed on the info up the 'chain', and hopefully someone somewhere in Comcast will see it and a lightbulb will go off over their head.

[Beck38]

Well, scratch what I said before (!). Took all day yesterday, but confirmed that all the transmission links (from my location/ISP through right to the last 'hop' going into Astraweb/US) are not the problem. It's either the last piece of fiber (Cogentco) going into Astraweb/US or the front end to the Astraweb plant itself.

We shall see if Astra can figure this out, or if it's the fiber provider, they 'wash their hands' or take some pro-active action. Again, it looks from my end to be timing slips that occur some 190minutes apart.

[nntpjunkie]

That is monumentally strange but kind of cool at the same time....maybe your bits are in a time warp...

[Beck38]

In the transmission world, it's not strange at all, except that someone (or more) somewhere simply isn't responding to the alarms or error logs that are staring them in the face. Astraweb should be hip deep in error logs from their front end plant, clearly showing that every 190 minutes they are generating a slew of block repeats. That would lead them into looking at the border router logs, which would lead them to the transmission system port those logs are pointing to. But the transmission company (tracert says Cogentco) is ignoring system alarms which would, again, lead them to look at the logs in their equipment right off.

Unless, of course, they are simply ignoring them.

[mjmacky]

Alright Beck, I see you've already figured out it was not your ISP. I'm having the exact same issue, and I'm most likely using a much older the exact same version of jbinup than as you. Also tested on the US server, non-SSL. I'm only using 4 connections.

[Beck38]

Yep. If you do a tracert to the Astraweb/US ports, the last hop should be over Cogentco (a large fiber outfit in Silicon Valley), and as I said this looks very, VERY, suspiciously like framing slips on a SONET fiber link, if not something else going on with the 'front-end' at Astraweb. But I'll bet it's the fiber link.

Astraweb hasn't responded to my trouble ticket from 2 days ago when they said that 'intermittent' problems couldn't be solved. This isn't 'intermittent', it's very regular. Like clockwork. Check your stuff. Check the vendor's supplying you. Do your job.

I might add, that using JBinUp means that all those errors are 'fixed' by utilizing that uploader; if someone is still using PowerPost or one of it's variants, this problem will be generating a ton of (hopefully with pars correctable) problems, but that is depended on how fast the upload is running. I run 'pretty fast' (it's almost as fast as the fastest available at my house, quite a bit slower than a couple miles down the road but them's the breaks).

The folks at Astraweb may (MAY) feel that keeping an eye on the fiber feeds into their facility is 'not their job'. If a backhoe was perched outside their building when they came to work in the morning, digging away in their 'front yard', they wouldn't pay it any nevermind. When the links went out a couple hours later, they couldn't/wouldn't 'connect the dots'.

As many have noted in the past with Astraweb, 'what a way to run a railroad'. But I'm not going to say that any other provider is any better (nzbmatrix, located outside of Toronto, CA, has had several outages due to 'backhoe fades' and their hosting facility doesn't have diverse routes on it's fiber feeds).

Here's hoping that maybe someone will notice that alarm (that's been silenced) blinking away at some point.

FYI, the ISP/Comcast part of this problem that got me off the track of 'tracking it down' a bit, was caused by Comcast deciding to upgrade the firmware in my cablemodem. In talking to them this last Friday, they could 'upgrade' the firmware, but upon finding out 'after the fact' that the vendor (and the equipment) couldn't 'downgrade' it's firmware 'over the network'. Okay, yet more software folks who can't program their way out the of proverbial 'wet paper bag'. So, they said they saw problems with that upgrade, and had to come in and physically do it.

Well, it didn't work exactly, the visiting guy saw some other problems, and ended up swapping the entire unit out, which caused yet more problems. A couple hours later, those looked fixed, and he shook his head and said 'try it out', and away things went. Looks good from my standpoint, and I'm trying some heavy/fast inbound/download transfers today, which is so far working 100%, so it looks okay. Cross my fingers.

[Beck38]

Astraweb closed my ticket, although they also said they are 'monitoring' things but that it's too 'irregular' to try and 'catch'.

What about every 190minutes like clockwork is 'irregular' exactly..? I think they simply don't want to be bothered.

[Hole69]

Have you run local hardware diagnostics - RAM, HDD? Another NIC?

[Beck38]

Has nothing to do with anything, other than the links/ports at Astraweb/US, that has been proven, read back on the thread.

It doesn't affect transfers FROM the server; if it did, they would probably be hip deep in complaints, only data going TO it.

[erebos]

I'm using the same latest version of Jbinup uploading with Astraweb. I have the same error every now and then. I haven't tried yet with Powerpost, which I'll do later. Ill report any logs regarding it.

Could you try with Powerpost?

[Beck38]

I could, but since PowerPost is 'non-error identifying or correcting', it would propagate those errors, and I really don't have the time to check that out at present. I might in a few days be able to do so, posting to binaries.test or something like that. You'll have to look at what you've posted, plus your error logs in PowerPost (are there any?) or else keep your eyeball on it during the 'period' when the errors occur (pretty easy to do).

One thing I have verified is the newsgroup I'm posting to; it makes no difference, got it on a couple different ones, as well as changing the route the data takes to Astraweb/US. From my standpoint, everything 'hits' Congentco as the last 'hop' going in, I wonder if anyone sees any other carrier as the final hop (doubt it, but...)

[johhny]

it happend to me to yenc power post,but only at very small files like samples,subs etc after few retrys it worked again.
using astraweb Europe Server non-ssl.

[Hypatia]

Beck, did you contact AW support?

[mjmacky]

Beck, did you contact AW support?

inb4becksaysreadthethread


Astraweb hasn't responded to my trouble ticket from 2 days ago when they said that 'intermittent' problems couldn't be solved. This isn't 'intermittent', it's very regular. Like clockwork. Check your stuff. Check the vendor's supplying you. Do your job.


Astraweb closed my ticket, although they also said they are 'monitoring' things but that it's too 'irregular' to try and 'catch'.

What about every 190minutes like clockwork is 'irregular' exactly..? I think they simply don't want to be bothered.

[Beck38]

Now going on 11+ days since I first noticed this. 8 days since I entered a support ticket, and 5 days since they closed it with saying they are 'monitoring' it. Errors continue like clockwork, ~190min between 'events'.

If those occur right at the end of a posting (that's been going on for x hours), trouble will ensue with JBinUp. You'll probably (I did) have to re-upload that last bit by 'restarting' it within JBinUp after the 'event' window has passed. This has now happened to be once, and it did restart and finish the upload, and allowed me to generate the nzb.

I've done a couple of rather large download transfers (4-5 hours/50GB worth) without any errors/problems, plan on doing another tomorrow, so like I told the Astraweb folks, if it were something do to with downloading, they'd be hip deep with complaints.

Here's a 'funny'; yesterday, a Comcast person who was canvasing my neighborhood/cul-de-sac trying to sell cable service, I chatted with at my front door for a bit. I 'avoid' the restrictions on consumer service (caps) by being on 'business class', and I mentioned that Comcast buys internet at less than 1cent per Gigabyte, and has a (resent upgrade from 250GB/Month) to 300GB 'cap', pays $3 for something they then turn around and sell for $50+. They do the same with video services, which is why I have had satellite (in one form or another) since 1982.

He was flabbergasted. In my cul-de-sac, out of some 18 homes only two have cabletv, everyone has satellite. Which is probably why he was given a 'check sheet' to go by of folks who don't subscribe to their 'service'.

Made my day.

[mjmacky]

That is strange, the most I've ever heard was Verizon trying to sell their service by phone, no boots on the ground.

[Beck38]

I live in a 'strange' neighborhood, in that it's surrounded on all sides with 'planned' communities with (illegal by federal law) restrictions on things like satellite dishes. The state atty. gen. backs those restrictions; I used to live in one of those before I retired, and I had to 'hide' my dishes behind tall fences and shrubbery, but could have been cited at any time by the property association.

But there is about 5-10 streets in the middle where there are no restrictions, and that's why almost everyone both on my street and several adjacent are filled with folks who like the little city we live in, like the view of the ocean and such, and have a high disgust of cable companies in general. The biggest downfall is no FIOS. All those communities surrounding have it, but not the little pocket we live in. Only Comcast. Of which a fair number (including myself) in our little slice of the city have internet service from. Like we got a 'choice'.

Now that state atty. general that allows these community associations to fine folks for having dishes (R), is running for Governor. You can bet who his biggest money backers are.

[mjmacky]

Oh please don't get me started on crooked association practices, I spent my load not a week ago.

A friend of my once had Fios set up in his apartment. There were three large boxes installed in 3 separate locations, besides his modem/router, I was quite baffled as to what each of their functions were.

[Beck38]

If anyone is still using PowerPost or one of their variants, with the current problems with Astraweb/US, you will be generating a nice little 'bad spot' in your upload, period. Verified. Hopefully, enough pars to correct all of those that get generated every 190m of your upload.

JBinUp 'sees' the bad spot, and corrects it, at least so far that I've been watching this.

[erebos]

If anyone is still using PowerPost or one of their variants, with the current problems with Astraweb/US, you will be generating a nice little 'bad spot' in your upload, period. Verified. Hopefully, enough pars to correct all of those that get generated every 190m of your upload.

JBinUp 'sees' the bad spot, and corrects it, at least so far that I've been watching this.
Dude.. thanks for the experiments...

[mjmacky]

Actually, if a lot of people are using powerpost, we should end up seeing a lot of incompletes over the past week.

[johhny]

astraweb/eu never had any issues for me so far with yenc

[mesaman]

Show me an example of AW or PowerPost upload failure.

[Beck38]

There are no 'failures' with JBinUp, since the program checks the server for completeness and when it finds the 'holes' created by the 'too small' errors it re-uploads the portion that was mangled and the propagation is complete/clean.

However, when using a non-error correcting uploader like PowerPost, the error is not 'cleaned up' and propagates out. PP doesn't generate any log files, and in fact 'pre-generates' the nzb before uploading, unlike JBinUp witch doesn't generate it until the complete files are upload verified. That's the 'magic' of JBinUp and why any errors one ever gets from something uploaded by it is from either bad server storage, bad server propagation, or bad transmission from that server to your system, not bad uploading.

The only way one using PowerPost (or variants) and catching the problem is either by watching the streams as they are uploaded, or by checking the nzb completeness either (or both) at the server you uploaded to or by one 'downstream' servers.

[Beck38]

At appox 0130hrs this morning (some 7+ hours previous to the date/time stamp on this message) the Asutraweb/US plant (or the front-end) took a major dump/reset. JBinUp sailed through it like it usually does, correcting things as it went along. Lots of 'connection resets' and the like.

After that, it appears that there are no more 'too small' errors occurring. It's only been some 7 hours since the 'reset', so it's only had a couple of 190min gaps to 're-appear', but it looks smooth as silk right now. Will of course continue to 'monitor..

More: Did get a message from the folks at Astraweb, they did make some changes and implemented them at that time (some 10+ hours ago now), and it seems to have 'cured' the 'too small' problems plus really smoothed out their operation; haven't even gotten a 'connection reset' message over those 10+ hours, VERY SMOOTH.

Hopefully it will continue this way, will continue to monitor (of course) but it looks really good right now!