ls: reading directory .: Input/output error

[GoogleCodeExporter]

What steps will reproduce the problem?
1. Install Ubuntu 10.04 from Ubuntu AMI on ec2;
2. Follow the wiki for s3fs;
3. Problem duplicated twice;

What is the expected output? What do you see instead?

The expected output post /usr/bin/s3fs <insert unique s3 bucket name> /mnt/s3 
is that the mount works and files are accessible;

What happens is a message shows up in tail -f /var/log/messages: "s3fs: init 
$Rev: 191 $" AND when one cd's into the s3/ directory and then performs and ls 
or tries to create a file an error is presented instead: 

root@host:/mnt/s3# ls -la
ls: reading directory .: Input/output error
total 0

What version of the product are you using? On what operating system?

I am using the s3fs-1.35.tar.gz install on ubuntu 10.04 LTS; I had to change 
the configure.ac to allow the installation or it error'd out on fuse 2.8.4 as 
ubuntu 10.04 LTS is patched and working effectively on 2.8.1 and they will not 
change the version on the ubuntu LTS image.

Please provide any additional information below.

If you need any more information please let me know.

suntzu

Original issue reported on code.google.com by semacken...@gmail.com on 22 Jan 2011 at 4:24

[GoogleCodeExporter]

s3fs-1.35 is not compatible with FUSE versions lower than 2.8.4

This is not an s3fs problem.

Original comment by dmoore4...@gmail.com on 22 Jan 2011 at 7:36

• Changed state: Invalid

[GoogleCodeExporter]

Another obvious problem (and probably a bigger one) is that "$Rev: 191 $" does 
not correspond to the 1.35 tar ball.  The revision for 1.35 in src/s3fs.cpp is 
304

% which s3fs

% s3fs --version

should give additional clues.

Original comment by dmoore4...@gmail.com on 22 Jan 2011 at 9:52

[GoogleCodeExporter]

I have installed Ubuntu 10.10 from Ubuntu AMI on ec2 + s3fs 1.35  and I'm 
having a similar problem. Maybe it's related.

$ ls -la
d?????????  ? ?    ?       ?                ? mnt

$ tail -f /var/log/messages
init $Rev: 304 $

$ which s3fs
/usr/bin/s3fs

$ s3fs --version
Amazon Simple Storage Service File System 1.35
Copyright (C) 2010 Randy Rizun <rrizun@gmail.com>
License GPL2: GNU GPL version 2 <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

Original comment by gratt...@gmail.com on 22 Jan 2011 at 11:09

[GoogleCodeExporter]

I'm suspecting something with your /etc/fstab or command line invocation at 
this point.

What is the line in your /etc/fstab or your command line?

Is the directory that you are trying to mount to empty?  In what directory did 
you run the "ls -la"?  If in / , then the mnt directory typically is not empty.

Try running s3fs in foreground mode from the command line with the -f option,  
the messages can be helpful.

Do a grep for s3fs in /var/log/messages and /var/log/syslog

I'm not denying that you are having an issue, but s3fs has been well tested on 
EC2 using a Ubuntu 10.10 AMI

Original comment by dmoore4...@gmail.com on 23 Jan 2011 at 4:24

[GoogleCodeExporter]

I get this error with buckets that have names containing capital letters.  
Despite being "allowed" to assign the name in AWS console, s3cmd has issues 
with them as well.  

Original comment by drexell0...@gmail.com on 31 Jan 2011 at 9:30

[GoogleCodeExporter]

I am having the same problem.  I am using 
http://blog.eberly.org/2008/10/27/how-i-automated-my-backups-to-amazon-s3-using-
rsync/ as a guide.

less /var/log/messages | grep s3fs
Feb  1 14:35:21 xxx-xxx s3fs: init $Rev: 304 $
less /var/log/syslog | grep s3fs
Feb  1 14:35:21 xxx-xxx s3fs: init $Rev: 304 $

$ s3fs --version
Amazon Simple Storage Service File System 1.35
Copyright (C) 2010 Randy Rizun <rrizun@gmail.com>
License GPL2: GNU GPL version 2 <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

$ which s3fs 
/usr/local/bin/s3fs

Original comment by mcpherson.rory on 1 Feb 2011 at 4:06

[GoogleCodeExporter]

I also get this problem with s3fs 1.40/rev 312 and libfuse 2.8.4 on Ubuntu:

# ls -lh
ls: cannot access mysqlbackup.sh: Input/output error
total 0
?????????? ? ? ? ?                ? mysqlbackup.sh

I suspect that some of this has to do with data generated by other s3 clients, 
as directories uploaded via s3cmd to reduced redundancy (and possibly others as 
well, I haven't tested that yet) aren't even showing up, and the the files I'm 
attempting to move into this directory I'm moving with s3cmd and were 
originally uploaded via s3cmd. S3 buckets were mounted with use_rrs=1 set.

Can data from other clients throw off s3fs this way, or am I barking up the 
wrong tree tree?

Original comment by joea...@gmail.com on 7 Apr 2011 at 5:38

[GoogleCodeExporter]

Data created by other S3 clients probably will not be compatible with s3fs. See 
issue #73, issue #21 and issue #31

Original comment by dmoore4...@gmail.com on 7 Apr 2011 at 2:25

[GoogleCodeExporter]

Ahhh, that makes sense about the difference concepts of directories. So, are 
there any workarounds, or is the best solution simply to re-upload data using 
rsync to my s3fs share?

Original comment by joea...@gmail.com on 8 Apr 2011 at 4:36

[GoogleCodeExporter]

There are workarounds, but tricky and easy to screw up. You'll need to use 
something like JetS3t Cockpit to add the directory objects, or you might be 
able to do it through the AWS console.  The simplest solution is as you 
suggested, but it might not be very fast if you have a slow internet connection 
or a lot of data.

Original comment by dmoore4...@gmail.com on 8 Apr 2011 at 4:46

[GoogleCodeExporter]

Cool, I don't mind reuploading the data.... Thanks!

Original comment by joea...@gmail.com on 8 Apr 2011 at 4:50