Am I getting longer running tasks ...
26 Jan 2020 1:20:51 UTC
Topic 220540
(moderation:
... or has my computer sprung a leak?
Until recently, the "Gravitational Wave Search" tasks ran in about 10 hours on my old Intel I3-based Linux box. The past week or so, I find some running for over 15 hours. The machine's done nothing but run 2 tasks at a time, each receiving 100% of a a thread on a 2-core 4-thread processor for more than a week. The processes are shown as each using 25% of the CPU.
The two running tasks show an estimated 15 hour running time. They're
h1_0186.00_O2C02Cl1In0__O2MD1C1_CasA_186.15Hz_13
h1_0186.00_O2C02Cl1In0__O2MD1C1_CasA_186.15Hz_12
Those tasks and all those waiting to run have an "Estimated Computation Size" of 144,000 GFLOPs.
Is it my machine or the universe at fault here?
The focus of the "search"
)
The focus of the "search" changed from O2MD1V1_VelaJr1 to O2MD1C1_CasA. That can be seen on the names of the tasks.
Different data... and these CasA tasks may require longer to crunch through. Personally all my six CasA tasks so far have ended up 'validate error' (linux host). i don't know yet if it's my computer or something else.
Thanks. Looking back, I now
)
Thanks. Looking back, I now see that the earlier tasks that ran long were actually "Gama Ray Pulsar something." One was LATeah1002F_1320.0_178994_0.0_0.
So none of the new ones have even completed on my machine. I have 2 running and 6 more in the queue. Looking at the elapsed + estimated remaining time for the 2 active tasks, it looks like 16 hours each now.
See you all (much) later, I guess.
All my O2MD1C1_CasA tasks
)
All my O2MD1C1_CasA tasks have gone straight to validate errors too. You are not alone. Probably the tasks are bad.
why should I spend so much
)
why should I spend so much computing time if so many were invalid.
bye bye einstein ...
The two tasks that finished
)
The two tasks that finished both failed with validate errors.
h1_0186.00_O2C02Cl1In0__O2MD1C1_CasA_186.15Hz_13_0
h1_0186.00_O2C02Cl1In0__O2MD1C1_CasA_186.15Hz_12_0
Looks like something is wrong.
At least it's keeping that part of the basement warm.
... that's an excellent
)
... that's an excellent question ...
maybe somebody can explain that to you?
When things like this happens
)
When things like this happens the project staff usually grants credit manually for failed tasks (validate errors) after the problem is fixed. So usually you'll get credit for the work done but the science is still lost.
San-Fernando-Valley wrote:...
)
Explain what, exactly?
Holmis wrote:When things like
)
Not worried so much about the credits, but about the repeated failures of the tasks.
If there's something wrong on my end, I'd like to fix it. If there's a problem with the software that's massaging the data, I'll stop taking new tasks until I see that it's been fixed. Or, if the error reports are actually spurious, I'll let things continue as they are.
Bert Hyman wrote:Holmis
)
I do share that worry.
If you look at your tasks and find that others also get validate errors then it will most probably not be your fault.
A validate error is declared when there is something obliviously wrong with the result returned, the validator does a sanity check of the result when returned before trying to compare it to your wingman, if that check fails then it's declared as a validate error.
If you run your gear out of default operating parameters and your wingmen returns good results, then it might be a good indication that your hardware is returning bad results.
If you get numerous validate erors and your wingman do too then it will probably be something wrong with the tasks, feel free to stop running them until the staff gets a chance to examine the problem.
Nothing wrong with reporting errors! That's one way the staff gets informed of what's happening!