On 04/05/2012 06:23 PM, Yunkai Zhang wrote: > I refactor zookeeper driver with following patches. > > These patches have been tested for several weeks, it can > support more than 400 sheep nodes. > > At this time, accord is not stable enough for us, this new > zookeeper driver maybe a good alternative. > > Yunkai Zhang (6): > Refactor zookeeper driver > Optimize the size of buffer send to zookeeper > try to recover previous zookeeper session when sheep start > retry again when zoo_* api return ZCONNECTIONLOSS/ZOPERATIONTIMEOUT > error > Fix two bug: > Use atomic builtins to replace pthread mutex locks > > sheep/cluster/zookeeper.c | 658 ++++++++++++++++++++++++++++++++++++--------- > 1 files changed, 524 insertions(+), 134 deletions(-) > Hi Yunkai, I think there is fatal problem in this patch set. Your first patch is being patched recursively by succeeding ones. We can only remove the code that are already in the tree. We should not try to remove the lines that are included by previous patches in the patch set. Thanks, Yuan |