When the disconnection is not caused by a broken node event, pick the state from a mapping: by default, a disconnected node is considered down (ie, removed from node manager as soon as it gets disconnected), with the exception of a storage node (it must be kept in partition table, so it must also be kept in node manager).


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1117 71dcc9de-d417-0410-9af5-da40c76e7ee4

Do not silently ignore an unknown node in disconnection handler method. Any node being disconnedted at the level of this handler must be known, otherwise it's a bug.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1116 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1115 71dcc9de-d417-0410-9af5-da40c76e7ee4

care about greater IDs received in service state, we don't have to receive this
packet.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1114 71dcc9de-d417-0410-9af5-da40c76e7ee4

use Dispatcher.unregister() to avoid duplicates.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1113 71dcc9de-d417-0410-9af5-da40c76e7ee4

Make count_dict property always contain all nodes, even if they are registered for the first time in FEEDING state.
This fixes the annoying KeyError raised in log method when it happens.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1112 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1111 71dcc9de-d417-0410-9af5-da40c76e7ee4

- Don't complain about deadlocks when reentrant (they cannot happen on a single lock, and more complex deadlocks are just not reported here)
- Display a complete traceback when a deadlock is detected (just displaying the line acquiring the lock is sometimes not enough).


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1110 71dcc9de-d417-0410-9af5-da40c76e7ee4

poll.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1109 71dcc9de-d417-0410-9af5-da40c76e7ee4

this must be already done, then if node object already exist, so create another
is wrong.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1108 71dcc9de-d417-0410-9af5-da40c76e7ee4

This fix the use of 'neoctl add all'.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1107 71dcc9de-d417-0410-9af5-da40c76e7ee4

message readability.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1106 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1105 71dcc9de-d417-0410-9af5-da40c76e7ee4

back without be rejected due to invalid server address.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1104 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1103 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1102 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1101 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1100 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1099 71dcc9de-d417-0410-9af5-da40c76e7ee4

Add a more detailed logging of ERROR packets (print error code and message).


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1098 71dcc9de-d417-0410-9af5-da40c76e7ee4

in event.py, now before trigger the event. It's thread-safe because the
connection is locked around the close().


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1097 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1096 71dcc9de-d417-0410-9af5-da40c76e7ee4

Rename partition_cell_states to cell_states to match with other names.
Fix indent and unify imports.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1095 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1094 71dcc9de-d417-0410-9af5-da40c76e7ee4

because there are Connection class atttributes.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1093 71dcc9de-d417-0410-9af5-da40c76e7ee4

Storage node register unknown storage node found in the database before loading
the content into the partition table. So an update() can now expect to known the
storage node.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1092 71dcc9de-d417-0410-9af5-da40c76e7ee4

called from the generic handler. This allow use it in some part of client and
storage handlers.
Add some XXX...


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1091 71dcc9de-d417-0410-9af5-da40c76e7ee4

- Renaming _nodeLost to _handleConnectionLost and removing the node parameter
whose the computation was done before each call to it.
- Renaming _dropIt to handleNodeLost, currently called only from the base
service handler, so the abstract definition as moved into it. In this case, a
call ensure that the node is not None.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1090 71dcc9de-d417-0410-9af5-da40c76e7ee4

call because the implementation do nothing. This is used only in
ClientEventHandler and so, no overloading is required.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1089 71dcc9de-d417-0410-9af5-da40c76e7ee4

ensure any acces to those methods can be done on any connection from the event
manager without crash the applications. This way is not less consistent than the
previous where isServerConnection was defined on ClientConnection and so on...
Replace any isinstance() calls with a connection as parameter by call to the
attributes above. 


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1088 71dcc9de-d417-0410-9af5-da40c76e7ee4

about the cluster state.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1087 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1086 71dcc9de-d417-0410-9af5-da40c76e7ee4

adn remove an XXX. Unify imports for protocol, remove unused variable.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1085 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1084 71dcc9de-d417-0410-9af5-da40c76e7ee4

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1083 71dcc9de-d417-0410-9af5-da40c76e7ee4

never send notifyNodeInformation packets to the master (and don't need to do
this), so remove the associated method handler on the master side and thus
remove an XXX.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1082 71dcc9de-d417-0410-9af5-da40c76e7ee4

ptid is different. 
Storage node no more use a 'ptid' attribute on the app instance but use the
getID() accessor from the PartitionTable class.
PartitionTable.clear() now also reset the ptid.
PartitionTable.load() do more checks (taken from master.recovery)


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1081 71dcc9de-d417-0410-9af5-da40c76e7ee4

Use max() instead of if <condition>: ...


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1080 71dcc9de-d417-0410-9af5-da40c76e7ee4

fix a potential bug. The ptid attribute on app is still removed.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1079 71dcc9de-d417-0410-9af5-da40c76e7ee4

master handler to ensure that the master connection is well established. Check
of greater ptid received with incremental partition table update is now done by
the PartitionTable class.


git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1078 71dcc9de-d417-0410-9af5-da40c76e7ee4