[sheepdog] [PATCH v4 0/5] Improving zookeeper driver

Kai Zhang kyle at zelin.io
Tue Jun 18 08:15:14 CEST 2013


v4:
- fixed compile errors
- update check_zk_rc() to handle ZOPERATIONTIMEOUT and ZCONNECTIONLOSS

v3:
- add another patch for handling session timeout of on the fly zookeeper
  operations

v2:
- fixed error of hang on joining cluster

Current zookeeper driver is fragile during bad network enviroment.
This series is to improve the robustness of zookeeper driver in fowlling ways:
- fix failure when multiple sheep join cluster concurrently
- re-join cluster when zookeeper session timeout
- handle ZINVALIDSTATE, don't call panic() simplely


Kai Zhang (5):
  zookeeper: fixed concurrent startup error
  sheep: rejoin cluster after a zookeeper session timeout
  tools: add kill_zk_session
  tests: add tests/065 for testing rejoin cluster
  zookeeper: handle session timeout for all zookeeper operations

 Makefile.am                |    2 +-
 configure.ac               |    3 +-
 sheep/cluster.h            |    1 +
 sheep/cluster/zookeeper.c  |  364 +++++++++++++++++++++++++++++++-------------
 sheep/group.c              |   12 ++
 tests/functional/065       |   46 ++++++
 tests/functional/065.out   |   38 +++++
 tests/functional/common.rc |   15 ++
 tests/functional/group     |    1 +
 tools/Makefile.am          |   13 ++
 tools/kill_zk_session.c    |   58 +++++++
 11 files changed, 448 insertions(+), 105 deletions(-)
 create mode 100755 tests/functional/065
 create mode 100644 tests/functional/065.out
 create mode 100644 tools/Makefile.am
 create mode 100644 tools/kill_zk_session.c

--
1.7.9.5




More information about the sheepdog mailing list