Hi everyone,

as my computer froze, I shut down the BOINC manager plus all tasks via the Windows task manager and restarted my laptop. After that, all my eight tasks were gone, the BOINC manager tells me there was a "calculation error".

All tasks were at around 25 % completion - frustrating for me as a new member since I'm eager to contribute as much time as possible.


Is there any way to reactivate those abandoned task?


Cheers
Philipp

There is no way for you to receive those tasks again. The tasks in those work units will be reissued to other computers.

It's frustrating to be sure, but it's happened to all of us at one time or another.

PhillipHoefl -

You might want to go to Tools -> Computing Preferences then "Under When to Suspend" and UNcheck all the boxes.

CPDN tasks don't like to be interrupted and it looks like yours are.

Also, check the box that shows "Keep non-GPU in memory when suspended".

Don't worry about your crashed tasks. Everyone gets them. Scientists can use the intermediate results (called trickles). If the scientists really need the complete result, it will be re-issued to another user.

---

Satisfying to know that at least my tasks are saved by the trickles (even tho I did not get any credits for them. Is that normal?)

However, I want to finish one of my tasks at some point - so I checked the entry of MikeMarsUK "Generic solutions to models crashing" to reduce as many crashes as possible.
Since that entry is from 2007 - are those suggestions still valid? What are your top 5 suggestions I could do for preventing crashes?

Thanks in advance, appreciate it!

Philipp

There's my post at the top of Preferences - here.

Trickle_up files don't always have science data included; most of it is in the zips that get created at about the same time.
These zips go back to a server "somewhere in the world", and are never see again except by the research people, so the only visible sign of an upload is the trickles listed on Tasks pages.

The credits for this project are different to all other projects - they are created by scripts using look up tables, based on the amount of processor effort that's required by different model types.
These scripts only run once per week, starting Saturday morning, UK time.
So you now have some.

Thanks again for the help. As you said, I got credits for the trickles today.

Next question: Does it make sense to stop all tasks I receive except for one. This single one can be calculated much faster in a shorter timeframe than four or eight tasks at once, thus I can contribute finished (and more "valuable"?) tasks much more safer. Since running four to eight tasks at once takes much longer, various technical faults could happen during that time, destoring valuable input, time and energy.

Does this makes sense or is your experience contradicting this?

Thanks again for answering my rookie questions!

Philipp

My preference is to not use hyper-threaded cores, as they don't add to much extra processing power for climate models.
If you follow this path, then it means that you have to know if your computer has them, and then set the BOINC option to 50% for that computer.

Which is under Computing preferences on your project's Account page, or the Options in your Manager's menu on your computer.

Also, as you have already downloaded several models, Suspend some of them rather than Abort some. Then run the others later.

******************************

BUT
Right now the project has storage problems, so NO zips are going to be uploaded for a while.

Best advice here is to Suspend all cpdn work until this is sorted out, hopefully early next week.

News item here.