vc: Restore sandbox state when there is error starting containers

[tedyu]

Fixes #2237

Signed-off-by: Ted Yu yuzhihong@gmail.com

[lifupan]

/test

[tedyu]

@grahamwhaley @jodh-intel
Please review

[lifupan]

/test

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (master@88205cf). Click here to learn what that means.
The diff coverage is 42.85%.

@@            Coverage Diff            @@
##             master    #2238   +/-   ##
=========================================
  Coverage          ?   49.06%           
=========================================
  Files             ?      111           
  Lines             ?    16063           
  Branches          ?        0           
=========================================
  Hits              ?     7881           
  Misses            ?     7207           
  Partials          ?      975

[tedyu]

@lifupan
Can you retrigger jenkins-ci-opensuse-15 test ?

Thanks

[grahamwhaley]

@tedyu - thought I should explain to you what that CI test fail is saying.... from the console logs:

bash sanity/check_sanity.sh
9b2373ad817eeb88d69666a02df2cba52ce5229159821b39591d6649fce815f1
ERROR: 1 pods left and found at /var/lib/vc/sbs
make: *** [Makefile:97: docker] Error 1

What the check_sanity.sh test does is ensure that after we've run tests, that the 'workspace' left by Kata is clean - that we have not left any processes or dirty files laying around on the system. In this case it found a file left in the sbs (sandbox?) directory that it did not expect to be there.

Letting you know just in case you need to debug - and your PRs are tinkering around that area of virtcontainers I think :-)

[tedyu]

@grahamwhaley
Can you re-trigger opensuse test run (to see if it was flaky) ?

thanks

[grahamwhaley]

ah, thought I had .... done now....

[tedyu]

@grahamwhaley @devimc
All tests passed.