Search code examples
bashdockernpmbackground-process

Bash: Multiple npm install in the background give error, 'No space left on device'

I am setting up docker on a google cloud compute machine, 1 vCPU and 3.75 GB ram.

If I simply run docker-compose up --build, it does work but the process is sequential and slow. So I am using this bash script so that I can build images in the background, and skip the usual sequential process.

command=$1
shift
jobsList=""
taskList[0]=""
i=0

#Replaces all the fluff with nothing, and we get our job Id
function getJobId(){
  echo "$(echo $STRING | sed s/^[^0-9]*// | sed s/[^0-9].*$//)"
}


for task in "$@"
do
  echo "Command is $command $task"
  docker-compose $command $task &> ${task}.text &
  lastJob=`getJobId $(jobs %%)`
  jobsList="$jobsList $lastJob"
  echo "jobsList is $jobsList"
  taskList[$i]="$command $task"
  i=$(($i + 1))
done

i=0
for job in $jobsList
do
  wait %$job
  echo "${taskList[$i]} completed with status $?"
  i=$(($i + 1))
done

and I use it in the following manner:

availableServices=$(docker-compose config --services)

while IFS='' read -r line || [[ -n "$line" ]]
do
  services+=$(echo "$line ")
done <<<"$availableServices"

./runInParallel.sh build $services

I string together available services in docker-compose.yml, and pass it to my script.

But the issue is eventually all the processes fail with the following error:

npm WARN tar ENOSPC: no space left on device, write
Unhandled rejection Error: ENOSPC: no space left on device, write

I checked inodes, and on /dev/sda1 only 44% were used.

Here's my output for the command df -h:

Filesystem      Size  Used Avail Use% Mounted on
udev            1.8G     0  1.8G   0% /dev
tmpfs           370M  892K  369M   1% /run
/dev/sda1       9.6G  9.1G  455M  96% /
tmpfs           1.9G     0  1.9G   0% /dev/shm
tmpfs           5.0M     0  5.0M   0% /run/lock
tmpfs           1.9G     0  1.9G   0% /sys/fs/cgroup
/dev/loop0       55M   55M     0 100% /snap/google-cloud-sdk/64
/dev/loop2       55M   55M     0 100% /snap/google-cloud-sdk/62
/dev/loop1       55M   55M     0 100% /snap/google-cloud-sdk/63
/dev/loop3       79M   79M     0 100% /snap/go/3095
/dev/loop5       89M   89M     0 100% /snap/core/5897
/dev/loop4       90M   90M     0 100% /snap/core/6130
/dev/loop6       90M   90M     0 100% /snap/core/6034
/dev/sda15      105M  3.6M  101M   4% /boot/efi
tmpfs           370M     0  370M   0% /run/user/1001

and here's the output for df -i

Filesystem      Inodes  IUsed  IFree IUse% Mounted on
udev            469499    385 469114    1% /dev
tmpfs           472727    592 472135    1% /run
/dev/sda1      1290240 636907 653333   50% /
tmpfs           472727      1 472726    1% /dev/shm
tmpfs           472727      8 472719    1% /run/lock
tmpfs           472727     18 472709    1% /sys/fs/cgroup
/dev/loop0       20782  20782      0  100% /snap/google-cloud-sdk/64
/dev/loop2       20680  20680      0  100% /snap/google-cloud-sdk/62
/dev/loop1       20738  20738      0  100% /snap/google-cloud-sdk/63
/dev/loop3        9417   9417      0  100% /snap/go/3095
/dev/loop5       12808  12808      0  100% /snap/core/5897
/dev/loop4       12810  12810      0  100% /snap/core/6130
/dev/loop6       12810  12810      0  100% /snap/core/6034
/dev/sda15           0      0      0     - /boot/efi
tmpfs           472727     10 472717    1% /run/user/1001
Solution

  • From your df -h output, root directory (/dev/sda1) has only 455MB free space.

    Whenever you run docker build, the docker-client (CLI) will send all the contents of the Dockerfile directory to docker-daemon which builds the image.

    So, for example, if you have three services each with 300MB directories, you can build them sequentially with 455MB available free space, but to build them all at the same time you need 300MB*3 amount of free space for docker-daemon to cache and build the images.