[upgrade] Upgrade failed with 'Sorry - this fdisk cannot handle AIX disk labels' in fuel_upgarde.log if run it with system test

[Andrey Sledzinskiy]

reproduced on upgarde script fuel-5.0.1-upgrade-23-2014-07-02_21-27-32.tar

Steps:
1. Install fuel 5.0 - ISO #26
2. Deploy simple cluster with all default values, 1 controller, 1 compute, 1 cinder
3. Upload upgrade script to master node to /var directory
4. Extract and run ./upgarde.sh

Actual result - upgrade fails with
start: (c,h,s) expected (%ld,%ld,%ld) found (%ld,%ld,%ld)
\00 ... %s ...
\00 Sorry - this fdisk cannot handle AIX disk labels.
 If you want to add DOS-type partitions, create
 a new empty DOS partition table first. (Use o.)
 WARNING: This will destroy the present disk contents.
\00 You have chosen an unusual boot file name.
\00 tv.tv_sec = %ld, tv.tv_usec = %ld
\00 tz.tz_minuteswest = %d
\00
 Be aware, that the bootfile is not checked for existence.
 SGI's default is "/unix" and for backup "/unix.save".
\00
 Bootfile is changed to "%s".
\00
 Bootfile must have a fully qualified pathname.
\00
 Name of Bootfile too long: 16 bytes maximum.
\00
 Sorry, no experts menu for SGI partition tables available.
full log is attached - fuel_upgarde.log
Problem occurs only if run upgarde with system test in which we assert that exit status from running ./upgarde.sh is equal to 0 but in real exit code is 255

After second run test also failed with unclear messages in log - fuel snapshot is attached and fuel_upgarde_second_run.log is attached

Manual run of upgarding helps and upgrade is successful

[Evgeniy L (rustyrobot)]

In the first log file, we can see part of file /usr/share/locale/et/LC_MESSAGES/util-linux-ng.mo which upgrade doesn't work with.
Also in the second file we can see another part of some random file.

As far as I know this happens only on system tests and it never happened if you run the upgarde script manually, am I right?

[Andrey Sledzinskiy (asledzinskiy)]

Yes, it only reproduced on system tests and only if test failed on some assertion step that could cause interruption of upgrade process

[Evgeniy L (rustyrobot)]

Andrey, does it still happen?

[Andrey Sledzinskiy (asledzinskiy)]

No, issue hasn't been reproduced last time

[Evgeniy L (rustyrobot)]

Andrey, it looks like some problem with tests, when it will happen again, could you please provide the env and also logs from system tests?

Will mark the bug as Incomplete for now.

[Andrey Sledzinskiy (asledzinskiy)]

Issue isn't reproduced