ClusterVerifyConfig can generate an error message for each instance.
These messages were pushed one-by one, triggering job reloading in Luxid.
In case of having hundreds of instances, this resulted in Luxid needing
>1GB of memory. This fix introduces ErroMsgList, that can report a list
of errors pushed in a single update. Also cleaned up the interface 
of erorr reporting, preparing for refactoring of the too generic 
Feedback(*args, **kwargs) into something more explicit and removing
of ErrorIf(bool, *args, **kwargs).



Viktor Bachraty (1):
  Fix ClusterVerifyConfig() causing high mem usage

 lib/cmdlib/cluster/verify.py                 | 79 ++++++++++++++++++++--------
 lib/jqueue/__init__.py                       | 32 +++++++++--
 lib/mcpu.py                                  |  1 +
 src/Ganeti/Constants.hs                      |  3 ++
 src/Ganeti/Types.hs                          |  1 +
 test/py/cmdlib/cmdlib_unittest.py            | 20 ++++++-
 test/py/cmdlib/testsupport/processor_mock.py | 12 ++++-
 7 files changed, 120 insertions(+), 28 deletions(-)

-- 
2.8.0.rc3.226.g39d4020

Reply via email to