bagri
bagri copied to clipboard
XML/Document DB on top of distributed cache
#156 hwo I can see my waif snapchat and whatts up chat @dexterDZ > @@``### 1.
shouldn't check cache size as it is too expensive: ``` 2018-06-18 13:52:58.609 [hz.Inventory-24.partition-operation.thread-51] WARN com.hazelcast.spi.impl.operationservice.impl.Invocation - [172.20.3.20]:10709 [Inventory] [3.10.1] Retrying invocation: Invocation{op=com.hazelcast.map.impl.operation.MapSizeOperation{serviceName='hz:impl:mapService', identityHash=1671141207, partitionId=627, replicaIndex=0, callId=600467, invocationTime=1529319178609 (2018-06-18 13:52:58.609), waitTimeout=-1,...
add methods to DocumentManagement interface: ``` T executeOnDocument(String uri, Callable f, Properties props) throws BagriException; ResultCursor executeOnDocuments(Collection uris, Callable f, Properties props) throws BagriException; ``` also can add two more...
currently we partition documents by uri hashcode. it is possible to add some special character to uri and use it for hash calculation. For instance: uri = uri123#4567doc. Here uri123...
Test failure is not always reproducing. > com.bagri.rest.service.QueryServiceTest#testQueryService assertEquals(3, results.size()); - actual result sometimes is 0, because of I/O error: `org.glassfish.jersey.server.ServerRuntime$Responder writeResponse SEVERE: An I/O error has occurred while writing...
- start Bagri node, run some long tests - in the middle of the tests start second node to join the first one - after node initialization and joining cluster,...
- should be taken into account on all client operations - can be implemented via configurable number of retries and delays between them - or via ReviableExecutionService
can be usefull when all operations on Documents can be done via DocumentManagement interface, no queries via QueryManagement interface, no indices also, having indices on certain fields used in queries...
must work for queries and documents
- close MBean connection server - close Rest server - close HZ node