1

I am mounting a S3 bucket on a amazon EC2 instance using S3FS. Things have been working great for now over 18 months.

Today I've tried to download a 5GB file using wget directly on the S3 bucket (things I already done before without issue) and I have a No space left on device error.

user@ec2-instance:$ /home/workspaces/$ wget http://example.org/file.xml.gz 
Cannot write to ‘files.xml.gz’ (No space left on device).

user@ec2-instance:$ df
Filesystem        1K-blocks    Used    Available Use% Mounted on
/dev/xvda1          8125880 6576804      1113264  86% /
none                      4       0            4   0% /sys/fs/cgroup
udev               15400288      12     15400276   1% /dev
tmpfs               3082424     216      3082208   1% /run
none                   5120       0         5120   0% /run/lock
none               15412116       0     15412116   0% /run/shm
none                 102400       0       102400   0% /run/user
s3fs           274877906944       0 274877906944   0% /home/workspaces

user@ec2-instance:$ df -i
Filesystem      Inodes  IUsed   IFree IUse% Mounted on
/dev/xvda1      524288 119569  404719   23% /
none           3853029      2 3853027    1% /sys/fs/cgroup
udev           3850072    380 3849692    1% /dev
tmpfs          3853029    262 3852767    1% /run
none           3853029      5 3853024    1% /run/lock
none           3853029      1 3853028    1% /run/shm
none           3853029      3 3853026    1% /run/user
s3fs                 0      0       0     - /home/workspaces

I don't understand what df -i return. I suspect that s3fs showing 0 inodes is an issue but I don't know what do do next.

This is not related to the Issue 288 in s3fs: No Space Left On Device for files larger than 5GB since:

  1. it is now fixed
  2. I've been able to download file larger than 5GB
Andrew Gaul
  • 262
  • 1
  • 7
magdmartin
  • 121
  • 1
  • 6

2 Answers2

3

According to this How to upload file directly to Amazon S3 from a remote server, you will have to download the file to local server before uploading it to S3 bucket.

Since you have not defined the save path, I afraid you are storing the large file locally at /tmp folder in which you don't have enough space for storing the 5 GB file.

You can further refer to this: How to upload a file to S3 without creating a temporary local file to see whether you can get rid of the temporary file created or expand your EBS volume to hold the large file before uploading to S3 bucket if you don't want to write an upload program.

Community
  • 1
Simon MC. Cheng
  • 436
  • 2
  • 7
  • This solved my problem. It also seems that s3fs isn't very good at cleaning up the /tmp directory. Trying to transfer a large directory full of files to an s3fs volume and it fills up /tmp and then stops working. Workaround seems to be to transfer in smaller chunks and clear /tmp in between – james-geldart Dec 21 '21 at 10:16
0

As Simon mention in his answer, the file is first downloaded locally at /tmp folder. Since my system disc has just 1.1GB available it was not enough to process the file.

To solve my issue, I expanded my ESB storage to get a larger /tmp folder.

proteus
  • 674
  • 5
  • 13
magdmartin
  • 121
  • 1
  • 6