0

I am trying to install tensorflow using conda in AWS EC2 instance for several times. But it is always giving below error. CondaMultiError: [Errno 28] No space left on device

I check df

(base) ubuntu@ip-172-31-23-129:~/anaconda2$ df
Filesystem     1K-blocks    Used Available Use% Mounted on
udev             5184264       0   5184264   0% /dev
tmpfs            1039260     720   1038540   1% /run
/dev/nvme0n1p1   8065444 7670180    378880  96% /
tmpfs            5196288       0   5196288   0% /dev/shm
tmpfs               5120       0      5120   0% /run/lock
tmpfs            5196288       0   5196288   0% /sys/fs/cgroup
/dev/loop0         93184   93184         0 100% /snap/core/6350
/dev/loop1         18432   18432         0 100% /snap/amazon-ssm-agent/930
tmpfs            1039256       0   1039256   0% /run/user/1000

Can anyone here help me in fixing this issue? Thanks!

Kuikiker
  • 156
  • 13
Anjuta
  • 13
  • 4

2 Answers2

1

(I will assume that you actually have space in the filesystem that you're installing to. In my case, I had space but the temp folder got filled up quickly, even though df -h and df -i did not suggest a problem. Since some of the filesystems in your list above are full or near-full, that may not be a good assumption. )

This answer might be useful for your needs, although I'm loathe to kill stuff in /tmp in case it's being used.

An answer that I like better is mkdir $HOME/tmp; export TMPDIR=$HOME/tmp I just used that one myself and it works just fine for my needs. Then you can delete the fake tmp directory after the installation works, and change TMPDIR back to /tmp

c74
  • 75
  • 4
0

Had simiar issue and solved it by resizinf tmpfs
You can resize it even during running conda env command

First, check the mount point for tmpfs:

> mount | grep tmpfs
...
tmpfs on /run/user/1000 type tmpfs (rw,nosuid,nodev,relatime,size=3229704k,nr_inodes=807426,mode=700,uid=1000,gid=1000,inode64)

Then resize:

> sudo mount -o remount,size=10G /run/user/1000
Slava Medvediev
  • 1,431
  • 19
  • 35