This patch series add the options to support the following work flow of online rolling reboots on a cluster with only redundant instances. First tag all nodes. Then repeatedly compute the next reboot group; this group of nodes safely can be drained and instances migrated off simultaneously. After rebooting, untagging, and undraining these nodes, other cluster operations can be done before continuing with the next group.
Klaus Aehlig (6): Fix expectation in hroller test Sort reboot groups by size Add option --one-step-only to hroller Support construction of the graph of all reboot constraints Support online-maintenance in hroller Add a test for online rolling reboot scheduling Makefile.am | 1 + man/hroller.rst | 28 ++++++++++++++++++++-------- src/Ganeti/HTools/CLI.hs | 21 +++++++++++++++++++++ src/Ganeti/HTools/Node.hs | 29 +++++++++++++++++++++++++++-- src/Ganeti/HTools/Program/Hroller.hs | 27 ++++++++++++++++++++++----- test/data/htools/hroller-online.data | 14 ++++++++++++++ test/hs/shelltests/htools-hroller.test | 14 +++++++++++++- 7 files changed, 118 insertions(+), 16 deletions(-) create mode 100644 test/data/htools/hroller-online.data -- 1.8.2.1
