runaway wu
16 Aug 2008 14:01:34 UTC
Topic 193863
(moderation:
I have a wu that has run for 15h and is 46% done; the estimate was 11h. So the runtime will be more than 2x the estimate, assuming it actually completes. S5r4a
Should I abort it, or is this normal behavior and so I should just let it finish?
runaway wu
)
You haven't been getting a mix of S5R3 and S5R4 work units have you, or recently started getting R4's? Just leave it unless it doesn't actually increase its percentage completed.
There is this message that tries to explain how BOINC estimates the time to complete a work unit and gets confused by R3 and R4's being mixed.
BOINC blog
I don't know how to get a mix
)
I don't know how to get a mix so I'm probably not. Thanks. I'll monitor the situation to see if the % increases.
RE: I don't know how to get
)
I just had a look at your hosts. Some are still working on R3 resends and some have transitioned to R4. I presume that this results list belonging to hostID 743164 will contain the task you are referring to because the two most recently completed tasks were R3 resends and the 4 current tasks are new R4 work. Your host would be working on the first of the R4 tasks which is showing the behaviour you describe. This is absolutely as expected and you should just let it run. The true crunch time could be 4 or 5 times the initial estimate. The DCF that applied for R3 work was probably around 0.3. The new DCF for R4 work will probably end up being of the order of 1.2 - 1.5 and your BOINC client can only "learn" this when the first R4 task is finished and the true crunch time is known.
MarkJ gave you a link where I've tried to explain all this.
Cheers,
Gary.