When an instance gets resized, but not rebooted afterwards, the state-of-record and state-of-world data differ. This can cause a discrepancy between 'hspace -L' predictions and hail behaviour, as the luxi backend receives state-of-world data, whereas the instance allocator is provided state-of-record data. Hence add a --state-of-record option to make htools ask for state-of-record data on the luxi backend as well.
Klaus Aehlig (4): Remove -X from hspace man page Add a new option --state-of-record Honor the --state-of-record option in the luxi backend Make hspace support --state-of-record man/hspace.rst | 8 +++++++- src/Ganeti/HTools/Backend/Luxi.hs | 33 ++++++++++++++++++--------------- src/Ganeti/HTools/CLI.hs | 10 ++++++++++ src/Ganeti/HTools/ExtLoader.hs | 2 +- src/Ganeti/HTools/Program/Hscan.hs | 2 +- src/Ganeti/HTools/Program/Hspace.hs | 1 + src/Ganeti/MaintD/Server.hs | 2 +- 7 files changed, 39 insertions(+), 19 deletions(-) -- 2.6.0.rc2.230.g3dd15c0
