ctdb: make client requests more reliable under load

[David Disseldorp]

The CTDB Linux-HA/Pacemaker resource agent uses the ctdb client utility
to monitor the status of ctdbd on a given node, it does this using the
"ctdb ping" command.

Under heavy load connections to the ctdbd unix domain socket may fail
with EAGAIN, subsequently the error is immediately returned to the
calling shell.

This fix is based against the 1.0.112b branch. Feedback welcome.

Cheers, David

--
The following changes since commit b4cb9a7163251828a1fd69a6ebb3ee370d7643e8:

  vacuum: close fd leak (2010-08-05 10:43:48 +1000)

are available in the git repository at:
  git://oss.sgi.com/ddiss/ctdb 112b-client-conn-retry

David Disseldorp (1):
      client: handle transient connection errors

 client/ctdb_client.c |   34 +++++++++++++++++++++++++++++-----
 1 files changed, 29 insertions(+), 5 deletions(-)