The JMX management console can be accessed at the following url
<Application_Server_Address>/rdf4j-server/management/
Halyard-specific MBeans can be found under the domain com.msd.gin.halyard.
Management bean for HBaseSail.
Every active repository will have one of these, as will every active HBaseFederatedService.
For instance, the following JMX object names
type=com.msd.gin.halyard.sail.HBaseSail,id=1bce0f03,table=mydata,federatedServiceResolver=471560dc
type=com.msd.gin.halyard.sail.HBaseSail,id=6a3e78a2,owner=471560dc,table=otherdata,federatedServiceResolver=325ac1b6
indicate that the first MBean is the Sail used by the Repository for the table mydata,
and the second MBean is the Sail used by the HBaseFederatedService associated to the first (owner points to the first's
federatedServiceResolver) when performing SERVICE calls to the table otherdata.
The number of active connections.
Timeout in seconds for each query evaluation, negative values mean no timeout.
Boolean option to use HalyardEvaluationStrategy instead of org.eclipse.rdf4j.query.algebra.evaluation.impl.StrictEvaluationStrategy.
Sorted oldest start time first. Only the last ten queries are available.
running
If the query is running or has finished.
startTimestamp
When the query was started.
endTimestamp
When the query completed.
connectionId
Each query has a connection ID. This is useful if you want to kill a query by using the operation killConnection.
The connection ID can be used to link a query to its HalyardEvaluationExecutor as well.
queryString
The query that was executed.
queryTree
The QueryModelNode tree before optimisation.
optimizedQueryTree
The QueryModelNode tree after optimisation.
-
resultSizeEstimate - estimate of how many results the node will return based on the stats stored at the time of optimization. It also takes into account what bindings will be available at that point in the query.
-
resultSizeActual - the true number of results returned. Requires trackResultSize to be true, and to track at all nodes (rather than just branch nodes), trackBranchOperatorsOnly to be false.
-
totalTimeActual - time taken in executing the node. Requires trackResultTime to be true, and to track at all nodes (rather than just branch nodes), trackBranchOperatorsOnly to be false.
-
costEstimate - cost of executing the node.
A query plan is executed:
- Top down
- For Joins, left node then right node.
Optional ElasticSearch settings.
HBase table name used to store the data.
If enabled, only do result size and time tracking for branch nodes.
Changing this will affect future queries, not currently running ones.
There is a performance penalty for disabling this.
By passing the id of a connection you are able to kill the running query. These are found on a query in the recentQueries section.
Management bean for HalyardEvaluationExecutor.
The connection ID can be used to link this back to a query in recentQueries.
The number of binding sets per second being feed into the query from HBase reads.
The number of binding sets per second being produced by the query.
Management bean for TrackingThreadPoolExecutor.
If the node ordering doesn't look sensible the first thing to check is that the stored stats are sensible.
Link to [stats docs](../../docs/tools.md#Halyard Stats).