Urgent: some wu have been withdrawn, please check yours
28 Jun 2005 16:18:16 UTC
Topic 189440
(moderation:
All E@H work units which begin with 'h1' (that's a lower case 'h') have been cancelled by the project. If your machine has any of these, stop them now as they are simply wasting cpu cycles.
Certainly, when it comes to returning them, they will fail to upload and fail to report, and trying to do so may slow down the process of getting other work.
This does not apply to wu which start H1 (capital H).
More in this thread
Hope nobody minds the cross posts, but thought it would be good to get the word round as fast as possible. If you want to add anything to the subject of the h1 wu, please say it in that thread, to keep everything together.
~~gravywavy
Urgent: some wu have been withdrawn, please check yours
)
Hey Gravy,
You might want to add that now Bruce is going to grant credit to those afflicted by this, it might not be in your best interests to abort a running WU if it is virtually complete anyway. If credit is your motive then let it continue and only abort those WUs that haven't started yet.
However if science is your motive, abort anyway as there'll be no benefit to science in completing any h1_nnnn stuff now.
Please don't interpret this as a criticism. I think what you posted was absolutely right. I'm just adding a bit more to the rapidly evolving saga :).
One small nitpick. The WUs upload and report fine. I saw that happen to one of mine so there's no problem there. I actually saw it get granted zero credit and that's when I aborted all h1 work on all boxes I could get my hands on. I didn't want to waste science and waste credit :). Now after seeing what Bruce has subsequently done, that zero that was there before is now showing as the claimed credit!!
So at least I can assure people that the action taken by Bruce is working :).
Cheers,
Gary.
RE: If credit is your
)
sorry Gary, but with respect I think you haven't realised just how clever Bruce has been here - look at this wu (not one of mine) and you will see that cancelled wu will get credit for what they have actually done.
If you abort using the abort button, or if you stop BOINC and steal its files, either way the result gets reported and you get credit for the exact crunching you have done.
So there is aborting the misbegotten wu as soon as possible should be credit neutral but science friendly. Even if a user is totally fixated on the credit, I assume they'd prefer to get that credit for something that is useful to the project than continue to get credit for wasted time.
~~gravywavy
RE: sorry Gary, but with
)
Hey, isn't that lawyerspeak for, "You're totally wrong you twit ...." :).
You are correct with the first part - the abort button (if you have one) will report the WU at the stage it's at and you will get the credit to that point. However you are wrong with saying that stopping BOINC and deleting has the same effect. My comments were specific to V4.19 which many people are still using and which doesn't have an abort button. I actually mentioned this in several other posts at the time but neglected to make the distinction in this particular thread.
When I stopped partially completed results and deleted the h1_nnnn large data file on a number of boxes running 4.19, the immediate effect on restarting BOINC was to zero out the elapsed cpu time and attempt to restart the result. Of course it couldn't immediately until it attempted to reget the missing large data file, which in some cases it was actually able to do, surprisingly enough.
After much stuffing around I eventually got the h1 stuff purged and was left with results on the webpage showing as client error with zero cpu time. No credit was claimed or granted for those. For some machines in remote locations that I couldn't get to, the h1 work seemed to self abort and I got partial credit for the actual time spent. In some cases the h1 results went to completion and I got full credit. All in all a very satisfactory outcome to a rather nasty situation. (ie I didn't have to travel to fix things, they sort of self fixed :). )
Absolutely, if you have an abort button. No argument with that. Things were happening very quickly and I was only trying to help people fully understand their options if they were running 4.19 and having difficulty (as I was) in getting rid of the h1 stuff. I posted quickly and didn't make it clear about 4.19.
Cheers,
Gary.
RE: Hey, isn't that
)
no, the lawyerspeak for that is "yes, you'd expect that of course, but interestingly enough it turns out not to be so"
;-)
oh, that's not what I expected, how interesting that it turns out to be so...
see my latest posting on the main h1_ thread for another way to abort wu on 4.19
~~gravywavy