Still no SETI-BOINC stats

[Spacehead]

... or are you able to crunch & just the stats aren't updating? Just curious.

 

[TAandy]

"August 14, 2004
We encourage everyone to run SETI@home Classic during our server outages. SETI@home Classic makes the same scientific contribution as SETI@home/BOINC.

August 13, 2004
All the data has been copied to the Snap device. Both attempts to bring the project online met with an unexpected NFS issue. Roughly ten minutes after bringing the project online the NFS daemon drops the mount point. Hats off to the Snap technical support team who have been helping us work through this issue. We are still investigating this issue and don't expect to be back online until the beginning of next week. In the meantime everyone have a great weekend."

I've still got 12 left on one box, other one's running the BOINC CPDN beta test until they're back up again.

 

[Spacehead]

Thanks Andy. I guess i should of checked the main page. All i've been looking at is the stats pages to see if anything is happening.

 

[TAandy]

"August 16, 2004
Currently the project is up, as we were unable to recreate the problems we were seeing on Friday. Now there are occasional dropouts, but nothing devastating. We will keep everything running for the time being, but may have to stop if problems arise.

August 16, 2004
Today we are continuing to work with Snap technical support to solve our mounting problems. In the meantime, you may notice the project being started/stopped for brief periods as part of our testing."


Got 49 on one box today, now 13 on the other

 

[JarrodH]

I think that I might stick with Seti Classic right now. I downloaded Boinc for my new computer last night and could not get work the entire night. (8+ hours)

 

[JarrodH]

Whoops. Double post.

 

[lane42]

it seems to be somewhat working. Just got a bunch of wu's on 3 machines

 

[TAandy]

I managed to download 49 units on the 17th, have 29 completed, but apparently due to the database crash any units downloaded after 11th are all useless!
This is getting ridiculous, I used to think D²OL was bad, but for a project that is supposed to be ready for a public release, :thumbsdown:

 

[RaySun2Be]

I had some WUs completed, but didn't send or get new ones. Then they disappeared, but I cannot tell if they were sent or some problem occured. Didn't get any new ones.

I couldn't stand to see a cruncher go idle, so I switched it back to SETI1. I'm going to keep it there until BOINC gets quite a bit more stable.

 

[Rattledagger]

My biggest problem was to successfully boot one of the machines that was turned off due to thunderstorm. Looks like it doesn't like cold-boot or hot-boot, but somewhere inbetween. :roll:

Getting a connection with the scheduling-server was surprisingly easy, and in short order got assigned probably a full cache. But for some reason the data-server was slow so used 7 hours to actually download all wu. Of course this can be due to my connection...

The validator is still off, and maybe they'll make sure everything is running as it should before they re-start the validator.
There was also some talk of alpha/beta-testing the new hierarchical directory structure before turning on the validator, and since the project has been down they most likely haven't done this yet...

 

[Unforgiven]

the people over at SETI are completely ruining the little credibility they have remaining! this project has been an absolute disaster since its inception and is like watching a mike tyson boxing match....you dont want to watch but you stay tuned in to see what stupid thing is going to be done next!! ive been a loyal SETI I producer for well over a year and have done some damn fine processing in that time, but to think that people are going to hold on for this garbage with SETI II while the berkeley crew tries to figure out problems is straight up ignorant and out of touch with reality! the project needs to be closed down, put back in beta testing mode and re-opened when the staff has a concept of what they are doing. sorry for the harsh words but this project is absolute trash and has ruined any allegience that i may have had with it! i know others feel this way too so i know im not alone :|

 

[Rattledagger]

?

Not sure wtf you're really talking about...

Maybe overlooked something, but the BOINC core client doesn't seem to have any serious problems. There's some reports that auto-suspending and such doesn't work if you've got venue-specific preferences, and some linux-users has problems with missing libraries and such. A work-around for the 1st is only have default preferences, and the 2nd is upgrade your linux-distribution...
The seti-application seems to crunch along as it should.

Since both the core client & application AFAIK only has some very minor problems, can't see any reason to move this back to beta...
Also, CPDN has official release 26. August 2004 at 00:01 UTC, further indicating there's only small problems in the BOINC core client, so little reason moving back to beta.


The BOINC (seti/astropulse) beta has roughly 10k users, most only crunching a couple wu for every new release, and AFAIK the whole beta was running trouble-free from 1 single server-machine...
The "live" project has more users, every user has more machines, and the users tries to crunch all the time. Only then getting much higher load than in beta did the bottlenecks show up, and Berkeley is fixing these as fast as possible.
Or in a little different wording, if beta little load so impossible to know if a fix really fixes the bottleneck, if live shows the bottlenecks...

Some of the problems they've had is:
1; too slow splitter, more splitters added. too slow transitioner, partial fix more transitioners added.
2; database I/O-limited, turning off displaying of pending credit and such decreases the load.
3; code changes squees a little more out of the database, but unfortunately a bug was added. Yes, on this point you can complain.
4; fastcgi refused to work with something that worked in alpha, and also works in "classic"...
5; too many files in upload-directory, work-around needed to disable the validator. new code is tested before being implemented in "live" project.
6; the new Snap-box lead to raid-corruption of the database; reverting to the backup taken before copying to new box since the Snap was up less than a day, and probably to make sure the database isn't corrupted.
7; every copy or fix-script or restore runs slower than expected, so takes longer than expected to be up again...

1 & 2 doesn't show up in beta, 3 is Berkeley-error so should have been tested, 4 was tested but still didn't work, 5 is still waiting for the fix being tested before validator being turned on again, 6 is unexpected hardware-problem, 7 everything going slower than expected is actually expected...

BTW, from a forum-post 15 days ago: "Both Classic and SETI/BOINC are processing the same number of work-units a day. Classic is just making a bunch of machines process the same data over and over again. We reached parity with Classic a week or two ago. That is when the servers are up."

So less than 2 months from launch the BOINC-version had overtaken "classic"... so sorry, I still don't see the "disaster" you're talking about...


From my point of view Berkeley knows what the problems are, but since the actual fix, turning off classic and use this hardware isn't currently an option, they're doing that they can with limited supply of hardware. Also, if you complains about they not testing fixes in #3, you can't complain of currently not getting any credit since they now IS testing the fixes.

 

[RaySun2Be]

1; too slow splitter, more splitters added. too slow transitioner, partial fix more transitioners added.
2; database I/O-limited, turning off displaying of pending credit and such decreases the load.

I'm sorry, but some of these issues should never have been issues. They should have been able to predict with quite some accuracy the transaction flows for the new system, since they've been testing it for some time, since they have had experience with SETI Classic as far as database and throughput to be able to estimate the loads for the new system.

It IS a mess IMHO, when I sit there for 3 weeks or more with the same amount of credit, you go to the website, and never know if you are going to be able to even get to your account and stats or not, you can't check what pending credits you might have, etc.

Behind the scenes, it may be going well for the SETI BOINC core processing, but from a users standpoint it sucks. Especially for a "production" release.

And I see they are going to reset SETI again.

So I think they should pull it back to BETA version until the whole package is more stable and USABLE.



I've pulled out of BOINC and gone back to SETI Classic for now, even if we are just reguritating prior results. At least I can tell my PCs are actually working, and can view the results of that work.

 

[Unforgiven]

Originally posted by: RaySun2Be

1; too slow splitter, more splitters added. too slow transitioner, partial fix more transitioners added.
2; database I/O-limited, turning off displaying of pending credit and such decreases the load.

I'm sorry, but some of these issues should never have been issues. They should have been able to predict with quite some accuracy the transaction flows for the new system, since they've been testing it for some time, since they have had experience with SETI Classic as far as database and throughput to be able to estimate the loads for the new system.

It IS a mess IMHO, when I sit there for 3 weeks or more with the same amount of credit, you go to the website, and never know if you are going to be able to even get to your account and stats or not, you can't check what pending credits you might have, etc.

Behind the scenes, it may be going well for the SETI BOINC core processing, but from a users standpoint it sucks. Especially for a "production" release.

And I see they are going to reset SETI again.

So I think they should pull it back to BETA version until the whole package is more stable and USABLE.

apparently RaySun2Be knows wtf im talking about without any problems! in general, people want 2 things.....1) a project where you can ACTUALLY download and upload results so your computer doesnt sit idle for days/weeks on end 2) when you DO get actual work to process, they want to be able to see the fruits of their labor! im not sure what your role with the SETI folks is Rattledagger and why you choose to defend them so religiously, but there are serious issues with this project and to plead ignorance to people being pissed off about this and to continually go to bat for them is beyond me. im not trying to be an a$$hole but im tired of the friggin excuses from the berkeley staff. they have taken a loyal following and are completely pushing them away from the SETI II project and are absolutely BLIND to this! :| i WANT to run this project, i WANT to contribute to a cause i beleive in, i WANT to help........but everytime i try to install this on my crunchers, i CANT GET ANY FRIGGIN WORK!!!!!!! what the hell is the point? tell me as a common user why i should bother with this project anymore?!?

 

[Rattledagger]

Well, after sitting through both the downtime then their line was cut, and 5? months in 2002 there it was impossible to connect since the bandwith from Berkeley was swamped, getting up-to-date-stats has never been my main point.

Let's see, you've got the "transition plan":
"If you have problems with the new version, you can uninstall it and reinstall the old 'Classic' version of SETI@home."
"Once SETI@home/BOINC is running smoothly and versions are available for most platforms, we will turn off the data server for SETI@home Classic."

And you've got the client, that clearly is marked as "beta, M1".
Using these 3 facts, and especially "running smoothly" is atleast by me interpreted as "some problems can happen"...


Not getting any work is of course a reason to complain, but atleast the last message wasn't decipherable that you really was complaining about, except things like "berkeley crew tries to figure out problems" and similar. My last post both pointed out the project isn't a disaster from the science-point, and they does know that they're doing... most of the time. Also the reason it must be listed as "officially active" and not "public beta", since as "public beta" it doesn't get enough load to know what problems to fix.

Till the bug was introduced, there was little problem downloading enough wu to cache through any of the small server-outages.
The bug of course is Berkeley's error, and just complain away about this.
But it's not really Berkeley's fault the Snap is delivered a week later than expected, and one of the 1st thing it does is giving raid-corruption... So instead of 2-3 days of moving the database to the new box it's been 2-3 weeks of moving/fixing/restoring...

As a by-product the validator is currently off, but it would in all cases be off while the rest is down. If they've had the opportunity to debug the new hierarchical directory structure they'll probably turn on the validator next week, but with all the outages it's not certain they've had the time/hardware to do this yet...


As for predicting server-load and similar, most likely they did some predictions, one of them was that #4, fastcgi, could be added after 2 days, but since it refused to work correctly the database got added strain for some weeks...
Someone also expected the current system could handle 3x the load then database and similar was optimized, maybe this is much too optimistic...

edit - f.cking net-connection. :frown:

As for reasons to still run BOINC, more effective, fairer crediting, cheat-free, multi-project-support, and currently also to load-test the servers and make sure they work then "classic" is turned off.
If you've got a permanent connection, just attack to seti and let BOINC run. To make sure the computer isn't running idle, apparently if you run the latest version of seti-driver and chooses lowest priority you'll crunch "classic" then can't get work from BOINC, and BOINC otherwise.

If you needs the instant-gratification of getting credit, seti will due to the new anti-cheating-code & fairer credit aways lag behind. In this case, wait a week and use CPDN instead, you'll here get credited for every trickle, 10802 timesteps.
Also, some stats-pages shows your total BOINC-credit regardless of score, so even if chooses another project you'll not need to start on 0 again.

 

[RaySun2Be]

Sorry, Rattledagger but we will just have to disagree it seems.

Remember this thread: Seti for BOINC "officially active"

Dated 06/22/2004

Edit, they've changed the text:

June 22, 2004
Welcome to the new SETI@home! The project is now officially active. We've reset the project by deleting all results and workunits, and setting all credits (user, team, host) to zero.

2 months later, still having problems, server up, server down, stats up, stats down. Granted, some have been out of their control, some haven't. But they have had many problems.

Now I've been pretty patient, but this has NOT been a smooth start from a users viewpoint.

"If you have problems with the new version, you can uninstall it and reinstall the old 'Classic' version of SETI@home."
"Once SETI@home/BOINC is running smoothly and versions are available for most platforms, we will turn off the data server for SETI@home Classic."

At the rate they are going, I'd be surprised if they turn off SETI Classic before the end of the year.

 

[Rattledagger]

Ah, you missed quoting my
"Of course, it can mean officially active for beta-testing... "

Seeing the client is "beta, M1" some problems isn't really unexpected.

Since "classic" is sitting on servers roughly 3x more powerful than available for BOINC, would be surprised if "pending credit" and similar is working before "classic" is turned off...


Hopefully the Snap doesn't puke again, so there will not be so long outages again... Still, expect many short outages for code-changes and similar, atleast till BOINC reaches M2...

 

[Rattledagger]

August 23, 2004
The validator is up and running and credit is now being calculated for all those backlogged results.