search

traitecoevo / mortality_bci

Tropical tree mortality model using Barro Colorado Island as a case study
Other
7 stars 3 forks source link

test models using docker #91

Closed jscamac closed 9 years ago

jscamac commented 9 years ago

docker container needs > 3GB memory to compile models.

Increasing to 6GB using docker-machine create --virtualbox-memory "6000" --driver virtual box mem6GB

Seems to fixed the issue. However, if additional workers are added the models begin to fail to compile again.

This could limit the machines we use on the cluster (i.e. we require machines with lots of RAM).

A possible solution would be to compile the models when we build the docker image.

FROM traitecoevo/mortality_bci
CMD Rscript -e '<code goes here>'

So we could have lots of different docker containers for different models, or lots of different models compiled into one docker container.

Updating the models is then pretty straightforward and doesn't require rebuilding the main container.

jscamac commented 9 years ago

Using the above. docker-machine create --virtualbox-memory "6000" --driver virtual box mem6GB We can successfully run a controller and 2 workers.

jscamac commented 9 years ago

Compiling the model locally (outside of docker) indicates that the compilation of the model uses ~600 mb. So not sure why we are running into virtual memory issues.