- 05 Aug, 2009 27 commits
-
-
Vincent Pelletier authored
Merge and simplify admin handler classes to hanle just 2 cases: expected answer, and the rest. This fixes cluster state notification handling (which used to only log a line because the hadler method was defined twice...). git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1202 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
in NeoCluster class. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1201 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1200 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1199 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1198 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
The existing code was exiting election loop too early, as there might not have been an identified connection to primary master (ask/answerPrimaryMaster is quicker than ask/answerPrimaryMaster + request/acceptNodeIdentification). The correct way is to stop waiting for connections to nodes other than primary master, and this can be done in answerPrimaryMaster. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1197 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1196 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1195 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
No need to cleanum NodeManager content when connection to primary master is closed, since we will anyway flush it when trying to reconnect to a primary master. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1194 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
Change primary master state to RUNNING, as we are indeed connected to it (and it won't notify of its own state). git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1193 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
Add a new Node state: UNKNOWN_STATE. In this state, nodes are supposed to be RUNNING, but there was no successful attemp to connect to them yet, nor any notification that they were running. This is especially useful in tests to avoid a stopped node to appear as RUNNING until receiving a notification that it is in fact DOWN. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1192 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
others test cases. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1191 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
processes. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1190 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
Make all election handler classes inherit from ElectionHandler: master will send a notification to all nodes that a new master appeared, which includes the new master itself, so it must expect a NotifyNodeInformation packet. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1189 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1188 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1187 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1186 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
Add extra checks when stopping primary master: now check that current primary master is None before expecting a new one to appear. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1185 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1184 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
Hence, don't add non-running nodes to unconnected_master_node_set before each election attempt. Preserve negotiating_master_node_set between attemps to reuse possibly pending connections. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1183 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1182 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
- Add expect* methods, to allow continuing as soon as a condition is encountered. - Define master node count as a constant. - Make getMasterNodeList able to filter on node state. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1181 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
case. Fix dirty work on getTwoIDs() and replace while loops by a simple check and two ordered UUIDs are needed. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1180 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
from NeoTestBase. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1179 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1178 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1177 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1176 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
- 04 Aug, 2009 13 commits
-
-
Grégory Wisniewski authored
executed each time a transaction is stored and only if the last OID has changed. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1175 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
When primary master node gets disconnected, use NodeManager.update method with DOWN_STATE, so that it gets removed. Clear node manager content and re-add all master node given in the configuration before trying to reconnect to primary master. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1174 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
Also log nodes known by address. Trigger a node manager log when it gets updated. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1173 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1172 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
PrimaryMasterHandler must not be applied to the listening connection, but only to the one remaining client connection (ie, the connection we established to the primary master node). All other connections (listening & server) must get IdentificationHandler (actualy only using handleAskPrimaryMaster, so a more restricted handler would better fit). git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1171 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Grégory Wisniewski authored
file. New masters will be registered by an admin and their config updated to ensure no more than one master could run concurrently when they restart without known each other. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1170 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
Make connection.py call connectionAccepted with the new connection as parameter. Change connectionAccepted prototype, as fewer parameters are needed now that "conn" contains them. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1169 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
Apply SecondaryMasterHandler to all established connections after having set the identificationHandler on the listening connection. This fixes primary master node not notifying secondary master node disconnections. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1168 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1167 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
Instead of stopping the whole cluster, only stop masters. This avoids disconnecting neoctl from admin process, as it will not reconnect (to be tested separately). git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1166 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1165 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
It is only important to raise a type error when other's type might be equal to EnumItem. In other cases, just return False. git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1164 71dcc9de-d417-0410-9af5-da40c76e7ee4
-
Vincent Pelletier authored
git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@1163 71dcc9de-d417-0410-9af5-da40c76e7ee4
-