Open ANDRO90 opened 8 years ago
This is likely because you are running out of space allocated for Docker. Try removing some docker images and containers and retrying.
Note: the below commands will remove ALL containers and images. If you want to only remove select ones, change the command appropriately.
# Delete all containers
docker rm $(docker ps -a -q)
# Delete all images
docker rmi $(docker images -q)
Thanks for the reply, but it didn't work. I even tried to enlarge the Base Device Size (sudo dockerd --storage-opt dm.basesize=30G) from 10G to 30G but the building process crashed at the very same point.
Hi,
I am getting the same error. After my build crashed I listed the currently running docker images and there were 24 of them, totalling over 60GB (my linux partition only has 60GB allocated to it).
How much space is required to build the gpu image?
Is there a way to reduce the number/size of docker images used in the build?
I am also facing same problem. fedora 24, 12GB RAM, 16GB Swap, no other app running.
[ 78%] Building NVCC (Device) object lib/THC/CMakeFiles/THC.dir/generated/THC_generated_THCTensorMathCompareInt.cu.o /home/pawan/torch/extra/cutorch/lib/THC/generated/../THCTensorInfo.cuh:276:0: fatal error: when writing output to : No space left on device } ^ compilation terminated. nvcc error : 'cicc' died due to signal 11 (Invalid memory reference) nvcc error : 'cicc' core dumped CMake Error at THC_generated_THCTensorSortLong.cu.o.cmake:267 (message): Error generating file /home/pawan/torch/extra/cutorch/build/lib/THC/CMakeFiles/THC.dir/generated/./THC_generated_THCTensorSortLong.cu.o
Sorry, running behind on issues tracking. Travelling now, will take a look next week and update.
An easy solution might be to just setup auto build in Docker Hub, so you can just download the image and don't have to build locally. I'll have to modularize the Dockerfiles to make that happen since Docker has CPU/time limits on the builds. Will do this in the next week. Apologies for the delay.
Hey all, So I am facing the same issue. How did you guys manage to fix this or is there a work around?
I never actually got this specific build working, but I ran into similar space issues with other Docker applications.
My solution was to move Docker's container storage directory (default is /var/lib/docker/ on Ubuntu) to another, larger partition.
I followed the symlink method described here: https://forums.docker.com/t/how-do-i-change-the-docker-image-installation-directory/1169
In case it is a problem with small /tmp
partition.
mkdir $HOME/tmp
export TMPDIR=$HOME/tmp
Then, don't forget to remove ~/tmp
I was able to fix this issue by clicking "Remove all data" I got to this pop-up by clicking on the docker icon at the top of my computer screen
The building of the image stopped because of this error, and I don't understand where it comes from, since I'm installing everything in a AWS p2 instance with 61GB of ram and 100GB of storage (only 15% of which occupied after the crash). Any guess? I was building the gpu image, after the execution of all the preliminary steps (Nvidia drivers and Docker)