Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | Merge pull request #5387 from ↵ | Tor Brede Vekterli | 2018-03-20 | 4 | -27/+98 |
|\ | | | | | | | | | vespa-engine/toregge/abort-operations-to-storage-node-going-down Toregge/abort operations to storage node going down | ||||
| * | Use std::move to avoid unneeded atomic ops. | Tor Egge | 2018-03-20 | 1 | -3/+6 |
| | | | | | | | | Add method to check if content node is up. | ||||
| * | Check that queued command will be aborted when storage node goes down. | Tor Egge | 2018-03-20 | 1 | -0/+42 |
| | | |||||
| * | Abort queued operations if storage node is down in new cluster state bundle. | Tor Egge | 2018-03-19 | 3 | -4/+21 |
| | | |||||
| * | Use ClusterStateBundle instead of ClusterState in | Tor Egge | 2018-03-19 | 2 | -23/+32 |
| | | | | | | | | ChangedBucketOwnershipHandler::OwnershipState. | ||||
* | | Merge pull request #5378 from vespa-engine/balder/no-more-thread-priorities | Tor Brede Vekterli | 2018-03-20 | 17 | -1120/+352 |
|\ \ | | | | | | | Balder/no more thread priorities | ||||
| * | | - Use num_threads directly as it has a sane default. | Henning Baldersheim | 2018-03-19 | 1 | -105/+55 |
| | | | | | | | | | | | | | | | - Use std::make_ insteda of explicit new. - std::shared_ptr can be assigned from std::unque_ptr. Avoid using release(). | ||||
| * | | Reduce vertical bloat | Henning Baldersheim | 2018-03-19 | 1 | -67/+32 |
| | | | |||||
| * | | Use new way of specifying threads. | Henning Baldersheim | 2018-03-19 | 1 | -2/+1 |
| | | | |||||
| * | | std::make_ instead of 'new'. | Henning Baldersheim | 2018-03-19 | 3 | -239/+95 |
| | | | | | | | | | | | | | | | | | | | | | Conflicts: storage/src/vespa/storage/persistence/persistencethread.cpp Resolve merge conflict with removed code. | ||||
| * | | Use simple num_threads. | Henning Baldersheim | 2018-03-19 | 1 | -1/+1 |
| | | | |||||
| * | | Clean out priority blocking stuff in PersistenceThread | Henning Baldersheim | 2018-03-19 | 14 | -543/+93 |
| | | | | | | | | | | | | | | | | | | | | | Conflicts: storage/src/tests/persistence/filestorage/filestormanagertest.cpp Resolve merge conflict. | ||||
| * | | Remove thread priority concept, has not added anything but complexity for a ↵ | Henning Baldersheim | 2018-03-19 | 12 | -182/+94 |
| |/ | | | | | | | very long time. | ||||
* / | Immediately send GetNodeState reply on "no more merges" pending edge | Tor Brede Vekterli | 2018-03-19 | 5 | -15/+128 |
|/ | | | | | | Lets the cluster controller update the derived bucket space states as quickly as possible when merges are done for the global space, without having to wait for the normal reply timeout period. | ||||
* | Merge pull request #5373 from vespa-engine/balder/remove-multioperation-only | Tor Brede Vekterli | 2018-03-19 | 37 | -1109/+34 |
|\ | | | | | Balder/remove multioperation only | ||||
| * | Default destructors. | Henning Baldersheim | 2018-03-19 | 2 | -12/+11 |
| | | |||||
| * | Remove DocumentList with children... | Henning Baldersheim | 2018-03-19 | 1 | -5/+0 |
| | | | | | | | | | | | | | | Conflicts: storage/src/tests/persistence/filestorage/filestormanagertest.cpp Resolve merge conflict. | ||||
| * | Remove MultiOperation | Henning Baldersheim | 2018-03-19 | 35 | -1092/+23 |
| | | | | | | | | | | | | | | | | Conflicts: storage/src/tests/persistence/filestorage/filestormanagertest.cpp storage/src/vespa/storage/persistence/persistencethread.cpp Fix merge conflict. | ||||
* | | Add total bucket counter to NodeMaintenanceStats. | Tor Egge | 2018-03-16 | 7 | -22/+54 |
|/ | |||||
* | Reduce merge busy wait time from 30 to 10 seconds | Tor Brede Vekterli | 2018-03-15 | 1 | -1/+1 |
| | | | | | Should speed up convergence in smaller clusters while still avoiding mass merge sends in larger clusters. | ||||
* | Immediately send GetNodeState replies when leaving recovery mode | Tor Brede Vekterli | 2018-03-14 | 5 | -4/+43 |
| | |||||
* | Expose StateManager method to immediately send GetNodeState replies | Tor Brede Vekterli | 2018-03-14 | 5 | -11/+50 |
| | |||||
* | Clean up of StateManager code, no logic changes | Tor Brede Vekterli | 2018-03-14 | 1 | -73/+59 |
| | |||||
* | Merge pull request #5320 from ↵ | Geir Storli | 2018-03-13 | 3 | -3/+42 |
|\ | | | | | | | | | vespa-engine/vekterli/ensure-missing-replicas-have-different-checksum-group-from-empty-replicas Missing replicas must be in a different checksum group than empty replicas | ||||
| * | Use more appropriate node indices in test | Tor Brede Vekterli | 2018-03-13 | 1 | -3/+3 |
| | | |||||
| * | Missing replicas must be in a different checksum group than empty replicas | Tor Brede Vekterli | 2018-03-13 | 3 | -3/+42 |
| | | | | | | | | | | | | | | Otherwise we risk merges with more than 16 nodes being scheduled for only the empty replicas and not actually include the missing replica nodes. This fixes #5313 | ||||
* | | Remove unused reporters from host info (cpu, disk, memory, network). | Geir Storli | 2018-03-13 | 17 | -586/+6 |
|/ | |||||
* | Merge pull request #5308 from ↵ | Geir Storli | 2018-03-13 | 8 | -9/+42 |
|\ | | | | | | | | | vespa-engine/toregge/add-clusterstate-to-content-bucket-space Add cluster state to content bucket space. | ||||
| * | Use deducted template arguments. | Tor Egge | 2018-03-13 | 1 | -4/+4 |
| | | |||||
| * | Add cluster state to content bucket space. | Tor Egge | 2018-03-13 | 8 | -7/+40 |
| | | |||||
* | | Remove LatencyStatisticsProvider and usage in PendingMessageTracker. | Geir Storli | 2018-03-13 | 8 | -375/+1 |
| | | |||||
* | | Remove never used per storage node ops latencies in host info. | Geir Storli | 2018-03-13 | 4 | -118/+19 |
|/ | |||||
* | Rename function to better match member variable it updates | Tor Brede Vekterli | 2018-03-09 | 2 | -3/+3 |
| | |||||
* | Must take mutex when updating bucket space statistics | Tor Brede Vekterli | 2018-03-09 | 1 | -0/+1 |
| | |||||
* | Invalidate accumulated bucket space stats on recovery mode entry | Tor Brede Vekterli | 2018-03-09 | 4 | -3/+85 |
| | | | | | | | Recovery mode is entered whenever a cluster state or distribution config change is detected. Invalidating the bucket space stats ensures that the distributor does not report stale statistics from a previous cluster state version to the cluster controller. | ||||
* | Factor out common code into new method to print xml tag for | Tor Egge | 2018-03-07 | 2 | -18/+23 |
| | | | | pending or delayed single bucket requests. | ||||
* | Report delayed bucket requests. | Tor Egge | 2018-03-07 | 1 | -0/+14 |
| | |||||
* | Must explicitly handle setdistributionstates method on reply processing | Tor Brede Vekterli | 2018-03-06 | 1 | -1/+1 |
| | | | | | | | Otherwise ends up sending unexpected return values to the cluster controller, which silently fails during response processing. End result is continuous transmission of cluster states, as the controller does not understand that the state has been ACKed by the node. | ||||
* | Remove unneeded include only used for debugging | Tor Brede Vekterli | 2018-03-06 | 1 | -2/+0 |
| | |||||
* | Add handling of v3 setDistributionStates RPC from cluster controller | Tor Brede Vekterli | 2018-03-06 | 12 | -29/+500 |
| | | | | | ClusterStateBundle is populated with derived states, but these are not yet exposed via the bucket space getter function. | ||||
* | Revert "Add cluster state RPC v3 support on content nodes" | Tor Brede Vekterli | 2018-03-06 | 12 | -498/+29 |
| | |||||
* | Remove unneeded include only used for debugging | Tor Brede Vekterli | 2018-03-05 | 1 | -2/+0 |
| | |||||
* | Add handling of v3 setDistributionStates RPC from cluster controller | Tor Brede Vekterli | 2018-03-05 | 12 | -29/+500 |
| | | | | | ClusterStateBundle is populated with derived states, but these are not yet exposed via the bucket space getter function. | ||||
* | Pass whole cluster state bundle to _newClusterStateBundle in | Tor Egge | 2018-02-26 | 1 | -1/+1 |
| | | | | PendingClusterState. | ||||
* | Enable cluster state bundle in distributor. | Tor Egge | 2018-02-26 | 19 | -96/+109 |
| | |||||
* | Use ClusterState from DistributorBucketSpace in PendingBucketSpaceDbTransition. | Tor Egge | 2018-02-26 | 2 | -3/+8 |
| | |||||
* | Simplify distributor state checkers unit test. | Tor Egge | 2018-02-26 | 1 | -2/+1 |
| | |||||
* | Use cluster state in distributor bucket space. | Tor Egge | 2018-02-26 | 12 | -23/+23 |
| | |||||
* | Populate cluster state in distributor bucket space. | Tor Egge | 2018-02-26 | 4 | -7/+34 |
| | |||||
* | Simplify distributor unit tests. | Tor Egge | 2018-02-23 | 17 | -81/+77 |
| |