DSE Graph cluster metrics
DSE Graph exposes different graph metrics as JMX MBeans. MBeans are exposed with JConsole or the command nodetool sjk mx.
Metrics concerning the tables used to store graph data is found under org.apache.cassandra.db.Tables for each graph_name and graph_name_system.
Metrics concerning the index table for graph data is found under org.apache.cassandra.db.IndexTables for each graph_name.
Further metrics can also be found under org.apache.cassandra.metrics.Table for each graph_name and graph_name_system.
Several of the metrics are exposed under com.datastax.bdp.graph:
graph_name
Each graph has metrics exposed with the following attributes:
-
adjacency-cache-hit: number of times the adjacency cache has had a cache hit
-
adjacency-cache-miss: number of times the adjacency cache has had a cache miss
-
graph-opened: number of times the graph has been opened
-
index-cache-hit: number of times the index cache has had a cache hit
-
index-cache-miss: number of times the index cache has had a cache miss
In addition, four JMX operations to compute the adjacency cache size, index cache size, clear a graph adjacency cache, or clear an index cache are available:
-
adjacency-cache.size: computes the adjacency cache size (number of entries)
-
adjacency-cache.clear: removes all entries from an adjacency cache
-
index-cache.size: computes the index cache size (number of entries)
-
index-cache.clear: removes all entries from an index cache
|
These JMX operations are not cluster-aware, and must be invoked on each node, as appropriate. |
GraphQueryThreads, GraphScheduledThreads, GraphSystemThreads, GremlinWorkerThreads
All graph thread metrics expose the following attributes:
-
ActiveCount
-
CompletedTaskCount
-
CorePoolSize
-
LargestPoolSize
-
MaximumPoolSize
-
Pending
-
QueueRemainingCapacity
-
TaskCount
GremlinServerPlugin
A few attributes are available for the Gremlin Server:
-
Active: Boolean value of whether DSE Graph is enabled
-
AnalyticsGraphServerIP: IP address of the server
-
Port: Port number of the Gremlin server
Statements
The Statement metrics expose different CQL Statement latency metrics through JMX located under com.datastax.bdp.graph.Statements.
The statements tracked are:
- Delete
-
CQL DELETE statement
- Update
-
CQL UPDATE statement
- LoggedBatch
-
CQL BATCH statement that is LOGGED, such as BEGIN LOGGED BATCH…
- UnloggedBatch
-
CQL BATCH statement that is UNLOGGED, such as BEING UNLOGGED BATCH…
- SchemaAltering
-
CQL statement that alters the schema, such as CREATE MATERIALIZED VIEW…
- SearchSchemaAltering
-
CQL statement that alters a search index, such as DROP SEARCH INDEX ON …
- SelectSearch
-
SELECT statement using a search index in the WHERE clause
- SelectSecondary
-
SELECT statement using a secondary index
- SelectView
-
SELECT statement using a materialized view table
- SelectTable
-
SELECT statement that uses no index, but only a primary key and optionally a clustering key for selection
All statements have the following attributes:
-
50thPercentile
-
75thPercentile
-
95thPercentile
-
98thPercentile
-
99thPercentile
-
999thPercentile
-
Count
-
DurationUnit
-
FifteenMinuteRate
-
FiveMinuteRate
-
OneMinuteRate
-
Max
-
Min
-
Mean
-
MeanRate
-
RateUnit
-
RecentValues
-
StdDev
