euca-bundle-vol does not create essential tmpfs mounts when bundling Ubuntu images

[Neil Soman]

One of the symptoms of this is that network interfaces are not properly ifup'ed on first boot.

To reproduce this bug (SRU validation):
------------------------------------------------------------
Right after starting an instance from a standard uec-image (64bit), I use the following procedure:

from your desktop:
scp -i mykey.prov ~/.euca/* ubuntu@<instance>:/tmp

in the instance:
# because of bug #457283
   #* edit fstab
   #* replace /dev/sdb by /dev/sda2 and on the same line ext3 by ext2
   #* mount /mnt
bucket=testbucket
prefix=test1
. /tmp/eucarc
if [ $(uname -m) = 'x86_64' ]; then
    arch=x86_64
else
    arch=i386
fi
sudo -E euca-bundle-vol -r $arch -d /mnt -p $prefix -u $EC2_USER_ID -k /tmp/euca2-admin-xxx-pk.pem -c /tmp/euca2-admin-xxx-cert.pem -s 2048 -e /mnt,/root/.ssh,/home/ubuntu/.ssh
euca-upload-bundle -b $bucket -m /mnt/$prefix.manifest.xml
euca-register $bucket/$prefix.manifest.xml

Start an instance of the new emi and wait until it is running.

From the CC
 * Try to ping it -> fail
 * Try to ssh to it -> fail

ec2-get-console-output ends with the following:
mount: mount point /var/run does not exist
mountall: mount /var/run [365] terminated with status 32
mountall: Filesystem could not be mounted: /var/run
mount: mount point /var/lock does not exist
mountall: mount /var/lock [366] terminated with status 32
mountall: Filesystem could not be mounted: /var/lock
mount: mount point /var/run does not exist
mountall: mount /var/run [367] terminated with status 32
mountall: Filesystem could not be mounted: /var/run
mount: mount point /var/lock does not exist
mountall: mount /var/lock [368] terminated with status 32
mountall: Filesystem could not be mounted: /var/lock
[ 4.965206] EXT3 FS on sda1, internal journal

[Neil Soman (neilsoman)]

------------------------------------------------------------
revno: 221
committer: Neil
branch nick: euca2ools-1.0
timestamp: Mon 2009-10-12 20:05:41 -0700
message:
  fixes #450044
------------------------------------------------------------

[Jason Rogers (juk80x)]

This problem still occurs in revno 231

[Jason Rogers (juk80x)]

Check out line 944.

942 if self.debug:
943 print 'Making essential directory %s' % mount_location
944 os.makedirs(dir_path)

 It is indented, and therefore will not work unless you are running in debugging mode. Which is probably why duplicate bugs keep appearing even though it works in testing.

[Neil Soman (neilsoman)]

Jason, good eye. Fixed.

thanks for noticing that.
neil

[Neil Soman (neilsoman)]

Forgot to add revno. It is 232.

[Mozg (andrei-arhont)]

Neil,

After building and installing 232 revision, I get the following error:

# euca-bundle-vol
Traceback (most recent call last):
  File "/usr/bin/euca-bundle-vol", line 36, in <module>
    from euca2ools import Euca2ool, FileValidationError, DirValidationError, CopyError, MetadataReadError, Util
ImportError: No module named euca2ools

Any idea what could be the problem?

Thanks

[Neil Soman (neilsoman)]

Mozg, looks like it did not correctly install. Did you run "sudo make install"?

neil

[Thierry Carrez (ttx)]

Neil: for the karmic SRU we need to cherrypick this fix. Would this be fixed by applying the patches from rev221 + rev232, or does it depend on other euca-bundle-vol fixes (like the make_essential_devs function introduced in rev207) ?

[Thierry Carrez (ttx)]

After discussing it with upstream, was considered not critical-enough to warrant a karmic SRU. Will be fixed in Lucid soon. Please reopen if you feel otherwise.

[Dustin Kirkland  (kirkland)]

Fixed upstream, merging.

[Launchpad Janitor (janitor)]

This bug was fixed in the package euca2ools - 1.1~bzr238-0ubuntu1

---------------
euca2ools (1.1~bzr238-0ubuntu1) lucid; urgency=low

  * Merge from upstream bzr revision; should fix:
    - LP: #497495 - parse dir path correctly in euca-delete-bundle
    - LP: #439366 - show kernel and ramdisk in euca-describe-images
    - LP: #495618 - show blk dev mappings in euca-describe-image-attribute
    - LP: #485106 - read credentials from config file
    - LP: #450044 - create tmpfs mounts in euca-bundle-vol
    - LP: #479823 - correctly handle userid in euca-bundle-vol
    - LP: #461301 - fix user data encoding
  * Also, created Ubuntu packaging branch, with a common root to the upstream
    project
 -- Dustin Kirkland <email address hidden> Tue, 05 Jan 2010 23:13:36 -0600

[cloudsuser (cloudsuser)]

I am looking for to execute the shell script through ssh. Getting error as File not found: ''

1) $ssh -i key.pem root@<host> -C "sh /root/myScript.sh"
Output: File not found: ''

2) $ssh -i key.pem root@<host> -C "/bin/sh ec2-create-volume --size 10 --availability-zone us-east-1a"
Output: File not found: ''

Note:
1) I have read write permission to my remote machine script myScript.sh
2) I can able to manually execute the command "ec2-create-volume --size 10 --availability-zone us-east-1a" successfully
If i want to execute the same with ssh via script, how i should procede. Please help me on this.

What may be the issues ?

[Scott Moser (smoser)]

@cloudsuser,
  Your issue is not likely to be related to this bug. If you are unable to resolve your issue, please open a new bug.
   As a hint, you may get further debugging info using "/bin/sh -x". I believe that you're hitting general issues with running a command with arguments remotely in via ssh.
   Please open a new bug.
   Thank you.

[graziano obertelli (graziano.obertelli)]

This issue seems related to euca2ools not eucalyptus. Reassigning it. And closing it since it has been released.