WCG is "up"... very limited work sent as test

[MindCrime]

from their faq forums.

"Hi everyone, we've seen some questions repeated among multiple volunteers that we'd like to address quickly.

For volunteers that did not receive any new workunits to crunch, we'd like to emphasize that only a limited amount of workunits were sent out as part of a testing phase. For those of you who were confused by our previous messaging, we're sorry.

About the URL change, please disregard this notice and do not update the URL, as the existing URL is correct.

Finally, the common error that stats "Server is out of disk space", all WCG servers have an excess of free space, as does the shared filesystem. We are investigating the cause of this error and will update when it is resolved. The bug reports are a great help and very much appreciated.

If you have any other questions, please leave them below and we will answer them as soon as we can.

- WCG Team"


I'm guessing it's close to full up soon.

[Al]

I managed to get 14 wus on one machine. 12 of them errored out in 1 second. The other 2 completed, uploaded and validated. They are still having some issues to work through.

[Al]

In the last 45 minutes I received 114 tasks. Unlike the previous 14 that errored, these seem to be running without any problem so far.

[zombie67]

Since I (we) can't change our project preferences, I am stuck at zero resource share. So I am not getting any work because I have other projects at greater than zero. Ah well.

[Al]

I wouldn't be too concerned about it at this point. Though you do get credits, no hours are being credited so you won't be making any progress on badge upgrades.

[MindCrime]

seems like boinc is only asking it for gpu work and not cpu work. and i guess that's what zombie is referring to? the device manager on wcg side isn't functioning?

[Al]

You're more than likely correct. Whatever your preferences were when the migration started is what your preferences are now. Though you can change and save your preferences at WCG, it ignores them and uses the old preferences. I was fortunate in that I was accepting OPN1 (cpu) at the start of migration, so I've managed to get over 1000 wus as they are released. Unfortunately, you get no credit for that in WCG credits or hours, but for some strange reason you still get Boinc credits at the stats sites. Not sure how that works. Possibly those will be credited at WCG at a later date, which would be great for me since it would put me over the 100 year mark on OPN.

[MindCrime]

I'm not an IT guy at all, and maybe this is average but if i had to opine, i'd say this wcg migration has been quite the fumble.

[Al]

I agree. They are now 2 months out from the initial restart date. I'm sure WCG not having a conforming Boinc website had something to do with that. It might have been better if IBM had stayed actively involved until everything was up and running correctly instead of dropping the support at whatever point they dropped it. It is obvious now that that point was too early.

[myshortpencil]

Does anyone know if WCG is sending out work? I got some in the early July test but nothing since. Also, I don't see a server status page -- probably part of the non-BOINC-conforming website that Al mentioned.

I hope everyone is happy, healthy and enjoying their summer.

[Al]

Does anyone know if WCG is sending out work? I got some in the early July test but nothing since. Also, I don't see a server status page -- probably part of the non-BOINC-conforming website that Al mentioned.

Hey MSP, I haven't been poking it for work for the last week or more. My guess would be probably some, but not much and you probably won't get any unless you run a tickler. They have never had a status page...would be a welcome addition.

[myshortpencil]

Thank you, Al :D

[Al]

Not much to update here. They are still in testing. More work is available, including releases of gpu work. You CAN now update your preferences. Alas, you still get Boinc credits at Free-DC and BoinStats, but no credit or hours increases show at WCG. I'm hoping that at least the hours are logged somewhere and can be added in later, but they have never said that will ever happen.

[MindCrime]

Al (Sphynx, right?) I know you have far more and far better gpus, And way better internet than me but how are you getting that much work from WCG? The transfer failures are so frequent that ive turned things down to 1 max concurrent transfer per project. That worked for a bit yesterday but i still had to baby it and hit "retry now" a lot. Today, however, I'm getting "internet access ok - project servers may be temporarily down" and chrome prevents me from accessing the website because of a security certificate issue.

update: seems they've fixed their security certificate issues, i was able to clear my transfer queue but it still took multiple retries. and that was just for a handful of WUs. I believe it's like 3 files per wu.

[Al]

Al (Sphynx, right?) I know you have far more and far better gpus, And way better internet than me but how are you getting that much work from WCG? The transfer failures are so frequent that ive turned things down to 1 max concurrent transfer per project. That worked for a bit yesterday but i still had to baby it and hit "retry now" a lot. Today, however, I'm getting "internet access ok - project servers may be temporarily down" and chrome prevents me from accessing the website because of a security certificate issue.

update: seems they've fixed their security certificate issues, i was able to clear my transfer queue but it still took multiple retries. and that was just for a handful of WUs. I believe it's like 3 files per wu.
Yesterday was an anomaly, so don't base anything you do on what happened when their security cert expired. Things at WCG are improving today.

As for the amount of work I process, well, as I've mentioned before, you need a tickler to update as frequently as possible and you need to be running multiple clients or running an app_config running multiple wus on each client. Multiple clients work best. Old gpus are not a problem, I've run multiple wus on 750ti's and gt1030's. I run 6 at a time on a gtx 1650 Super. It's all about how many clients you run and how often that number allows you to ask for work.

[MindCrime]

i'm running on a 750ti and 2x 7970s. my issue is the transfers fail and timeout and get a pushback time. I'm running BPU to get work every 3 mins but i don't have anything that will retry failed downloads. does your "tickler" do that?

[Al]

i'm running on a 750ti and 2x 7970s. my issue is the transfers fail and timeout and get a pushback time. I'm running BPU to get work every 3 mins but i don't have anything that will retry failed downloads. does your "tickler" do that?
No, it won't retry stuck downloads. Historically, I've retried mine manually.

[MindCrime]

No, it won't retry stuck downloads. Historically, I've retried mine manually.

but you can retry all hosts remotely, right?

[Al]

Yes, I use BoincTasks.

[zombie67]

Wait, a tickler/updater CAN retry uploads/downloads.


--network_available
Tell the client to retry deferred network communication.


Something like:


C:\Program Files\BOINC\boinccmd" --host 192.168.x.xxx --passwd xxxx --network_available

So stick that in your tickler file to run every 5 minutes or so, to prevent the increasing back-off.