Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Server computer freezes when COMSOL hogs all available memory
Posted Jun 9, 2010, 1:59 p.m. EDT 3 Replies
Please login with a confirmed email address before reporting spam
Hello all,
I am running into a problem while running a large simulation in server-client mode. When COMSOL consumes all available memory on the server, the server machine becomes completely unresponsive. I cannot kill the "java" process from either the client or the server. This is obviously most annoying as I then have to do a hard reboot on the server. I don't understand why the Mac OS allows this and doesn't give an 'out of memory' error.
Both the client and the server machines are Macs. The server is a MacPro with 16 GB RAM running Mac OS X 10.5.8 (Leopard).
Any suggestions?
Cheers,
Misha Monder
I am running into a problem while running a large simulation in server-client mode. When COMSOL consumes all available memory on the server, the server machine becomes completely unresponsive. I cannot kill the "java" process from either the client or the server. This is obviously most annoying as I then have to do a hard reboot on the server. I don't understand why the Mac OS allows this and doesn't give an 'out of memory' error.
Both the client and the server machines are Macs. The server is a MacPro with 16 GB RAM running Mac OS X 10.5.8 (Leopard).
Any suggestions?
Cheers,
Misha Monder
3 Replies Last Post Jun 12, 2010, 5:27 a.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
I was running comsol from a computer cluster and had similar problems. I was trying solving something very complex and it was using 90 % of the cluster memory. It took very long to calculate one iteration, and it slowed also all the other processes on the cluster. We didn't find out how to limit the memory that comsol can use, so I had to change the model or the solver.
I was doing a reaction model that has many components in it. The solution for the memory usage was to use segregated solver and to put all the component consentrations in their own groups to the solver.
I was running comsol from a computer cluster and had similar problems. I was trying solving something very complex and it was using 90 % of the cluster memory. It took very long to calculate one iteration, and it slowed also all the other processes on the cluster. We didn't find out how to limit the memory that comsol can use, so I had to change the model or the solver.
I was doing a reaction model that has many components in it. The solution for the memory usage was to use segregated solver and to put all the component consentrations in their own groups to the solver.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thanks for your post Jaana.
What frustrated me was that I could not even kill the 'java' process on the server because it became completely unresponsive. Were you able to kill the process from a terminal?
We are taking the same approach as yours by using the segregated solver. However for our system, the segregated solver cannot solve beyond a certain parameter value. We are solving a highly coupled nonlinear system of equations that describe the physics in a fuel cell.
Cheers,
Misha
What frustrated me was that I could not even kill the 'java' process on the server because it became completely unresponsive. Were you able to kill the process from a terminal?
We are taking the same approach as yours by using the segregated solver. However for our system, the segregated solver cannot solve beyond a certain parameter value. We are solving a highly coupled nonlinear system of equations that describe the physics in a fuel cell.
Cheers,
Misha
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
even on a single (multiple CPU) its difficultto abort a run particularly in V4, specially if there are several sequences such as a parametric sweep or segregated solvers looped in, it seems that it runs through all steps anyhow once one hit a "stop", but this can take ages on large runs, and cancel is not better if not worse
I hope they will fix this by having a better abort flag/semafor send around in the next releases.
These are typically things I send to support, as suggestions for "improvements"
Have fun Comsoling
Ivar
even on a single (multiple CPU) its difficultto abort a run particularly in V4, specially if there are several sequences such as a parametric sweep or segregated solvers looped in, it seems that it runs through all steps anyhow once one hit a "stop", but this can take ages on large runs, and cancel is not better if not worse
I hope they will fix this by having a better abort flag/semafor send around in the next releases.
These are typically things I send to support, as suggestions for "improvements"
Have fun Comsoling
Ivar