AnswerPrimaryMaster is now expected to happen before NodeIdentification happens, so this test function becomes pointless.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@719 71dcc9de-d417-0410-9af5-da40c76e7ee4

Fix test_unknownNewPrimaryAnswerPrimaryMaster test (getNodeByUUID is called just once, so it retrieved the value intended for the check of answering peer connection).


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@718 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@717 71dcc9de-d417-0410-9af5-da40c76e7ee4

- add trying_master_node where needed
- decrease the expected number of getNodeByUUID calls by one, since the first lookup is rendered useless by splitting handlers.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@716 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@715 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@714 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@713 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@712 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@711 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@710 71dcc9de-d417-0410-9af5-da40c76e7ee4

- it does not register new primary nodes (this only needs to be implemented in the handler as primary node will always send all nodes it knows)
- it sets app.trying_master_node to None when connection fails


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@709 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@708 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@707 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@706 71dcc9de-d417-0410-9af5-da40c76e7ee4

only send START_OPERATION message is master is in RUNNING state


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@705 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@704 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@703 71dcc9de-d417-0410-9af5-da40c76e7ee4

verification stage since it's unexpected because the master already got a
partition table it trust and do not ask another.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@702 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@701 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@700 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@699 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@698 71dcc9de-d417-0410-9af5-da40c76e7ee4

Implement message queue. This change makes poll handle at most one packet per call, instead of handling all pending packets at once.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@697 71dcc9de-d417-0410-9af5-da40c76e7ee4

Only catch KeyErrors generated when fetching a connection to avoid hiding other exceptions. (missed in r687)


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@696 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@695 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@694 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@693 71dcc9de-d417-0410-9af5-da40c76e7ee4

pass msg_id to register so that we can answer to neoctl request with
right id


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@692 71dcc9de-d417-0410-9af5-da40c76e7ee4

'SecondaryEventHandler', used in the secondary role, is now renamed as
PrimaryMasterEventHandler since a secondary is only connected to the primary. On
the other side, the primary apply the SecondaryMasterEventHandler on it's master
connections.
Factorize code in connection(Closed|Timeout)/peerBroken and avoid some checks of 
conditions that can be assume in specialized handlers.
Clean imports and remove some instance checks.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@691 71dcc9de-d417-0410-9af5-da40c76e7ee4

- one for bootstrap
- one for monitoring
- one for answer to request
add a dispatcher class to manage messages


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@690 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@689 71dcc9de-d417-0410-9af5-da40c76e7ee4

Send StartOperation packet to running storage nodes only.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@688 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@687 71dcc9de-d417-0410-9af5-da40c76e7ee4

- All incoming connections start with the identification handler which will
apply the right handler depending on the current cluster state and the node
type. So we are now sure that a node is well identified, got an UUID and is
registered into the node manager when it leaves the identification handler.
- During service, clietnt and storage nodes have their own handler


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@686 71dcc9de-d417-0410-9af5-da40c76e7ee4

exception during packet decoding or processing.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@685 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@684 71dcc9de-d417-0410-9af5-da40c76e7ee4

one storage node.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@683 71dcc9de-d417-0410-9af5-da40c76e7ee4

the current cluster state, so the primary master should not change the admin's 
connection handler when doing a state-switch.
Factorise a little the response to an identification request.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@682 71dcc9de-d417-0410-9af5-da40c76e7ee4

- first, asks which master is the primary
- then, when connected to the primary, request node identification
- then, ask explicitely for partition table and node information, to follow query/answer schema.
Details:
- Master:
 - Move handleAskPrimaryMaster, handleAskNodeInformation and handleAskPartitionTable implementations to base handler to factorise code
 - handleRequestNodeIdentification is not required any more in secondary mode, only handleAskPrimaryMaster
- Client:
 - Make bootstrap handler and master connection behave more similarly to on storage's equivalents
 - Remove PrimaryHandler (moved into PrimaryNotificationsHandler)
 - parse master node configured ip list just once, and register them to node manager just once (in app's __init__)


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@681 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@680 71dcc9de-d417-0410-9af5-da40c76e7ee4