[Sheepdog] [PATCH 00/14] fix I/O errors while node membership changes

MORITA Kazutaka morita.kazutaka at lab.ntt.co.jp
Sat May 14 09:03:45 CEST 2011


This patchset includes some cleanups and bug fixes related to node
membership change.

If this patchset works well, I'd like to release the version 0.2.3.


MORITA Kazutaka (14):
  use more suitable type
  introduce object recovery state
  fix I/O accesses to multiple unrecovered objects
  check return value of get_nth_node()
  avoid calling update_cluster_info() in worker threads
  fix reading responses of forwarded write requests
  set return value in all cases
  retry when network error occurs
  check epoch when the request causes local access
  fix wrong call of setup_access_to_local_objects()
  return SD_RES_NEW_NODE_VER during node membership change
  return SD_RES_NEW_NODE_VER when accessing recovering objects
  remove failed node from cpg ring
  retry getting object list when an error occurs

 include/sheep.h    |   12 +++-
 sheep/group.c      |  134 ++++++++++++++++++++++++++++++++++++++------
 sheep/sdnet.c      |   36 +++++++++++-
 sheep/sheep_priv.h |    4 +
 sheep/store.c      |  156 +++++++++++++++++++++++++++++++++-------------------
 5 files changed, 262 insertions(+), 80 deletions(-)




More information about the sheepdog mailing list