New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
MultiApp restart bug #5695
Comments
I'll take a look at this later tonight... unfortunately I'm tied up all afternoon. Thanks for the simple example and the explanation |
Thanks. Also, I have not seen a test for multi-app restart (maybe I did not look close enough), so this could serve as a basis for it... |
Ok - I see the issue. It's a restart vs. recover issue again. The sub-app is being "recovered"... even though the master app is being "restarted". The problem is that the sub-app gets its exact state pushed down to it from the master app. That's the right thing to do for "recover" but wrong for "restart" (the difference is in the While I think about this a bit... you may need to just use "recover" for the sprint problem. I think I have a fix... but I need to stew on it for a while... |
Actually - I don't think this will be that bad... stay tuned |
As a developer I'm a little confused about when I should use one or the other of these methods. Is there guidelines for how to talk to users about this? I guess users have to ask the question do I want the state to start over for "restart" but not for "recover"? Is that your thinking? |
@permcody For our users it's easy: they should only ever use
Easiest one to think about is However... during Restart Therefore So to summarize: Restartable means: state is kept for both Restart and Recover Hope that helps. |
This is incredibly nasty. It's 2AM here and I'm calling off the search until tomorrow. I think I need to take a different approach... |
Sorry for asking here. Actually restart/recover also confuse me. It is easy to understand why we have recover. But not very for restart, why do we just start the simulation from scratch, what is the difference? |
Restart is much like the two-part simulations you run. You might use data On Mon, Sep 21, 2015 at 8:57 AM Yaqi notifications@github.com wrote:
|
Oh thanks. Now makes sense. But is restart a subset of recover? If so we should just let user to make the difference. |
Restart is a subset of recover and I really don't think we want the user Cody On Mon, Sep 21, 2015 at 9:19 AM Yaqi notifications@github.com wrote:
|
Yep. For now, I'm managing access to All the other systems in MOOSE should never even be aware of it. |
@friedmud, you may want to check this out. It appears that we still have a few weird restart startup issues I had to remove the 'initial' execute for the postprocessors in this test to maintain the current gold file. refs idaholab#5695
Here is a branch that shows a bug in multi-app restart.
The problem being solved is very trivial. The sub-app is computing
u = x * t
on a 1D domain fromt = 0..1
withdt = 0.1
. This solution is transferred onto master app into a variable calledv
. The master app computesu = x * t
on a 2D domain.For the first part, I let the
t = 0..0.5
(i.e. 5 time steps) withcheckpoint = true
. Then the second part restarts fromt = 5s
and goes tot = 10
(another 5 time steps). While all is good in the master app, we end up with 5 time steps in the output file and both solutions (u
andv
) are correct, we have a bug in the sub-app:0, 6, 7, 8, 9, 10
(master app times steps are0, 1, 2, 3, 4, 5
) - from the console.ncdump
, because paraview was showing all zeros and incorrect times (because thetime_whole
var in the file reads5, 0, 0, 0, 0, 0, 0.6, 0.7, 0.8, 0.9, 1
, while it should be0.5, 0.6, 0.7, 0.8, 0.9, 1
).The text was updated successfully, but these errors were encountered: