Version
ExecuteGraph
Jobflow Component
Short description
ExecuteGraph runs graphs with user-specified settings and provides execution results and tracking details to output ports.
| Same input metadata | Sorted inputs | Inputs | Outputs | Each to all outputs | Java | CTL | Auto-propagated metadata |
|---|---|---|---|---|---|---|---|
⨯ |
⨯ |
0-1 |
0-2 |
✓ |
⨯ |
✓ |
✓ |
Ports
| Port type | Number | Required | Description | Metadata |
|---|---|---|---|---|
Input |
0 |
⨯ |
Input tokens with graph execution settings. |
Any |
Output |
0 |
⨯ |
Execution information for successful graphs. |
Any |
1 |
⨯ |
Execution information for unsuccessful graphs. |
Any |
The component can have a single input port and two output ports attached.
Metadata
ExecuteGraph does not propagate metadata from left to right or from right to left. It propagates metadata of its templates.
This component has metadata templates on all its ports. The templates are described in Details. See details on metadata templates.
ExecuteGraph attributes
| Attribute | Req | Description | Possible values |
|---|---|---|---|
Basic |
|||
Graph URL |
yes |
A path to an executed graph.
Only a single graph can be specified in this attribute.
The value can be overridden by a value from input token, see the |
|
Execution type |
no |
Specifies a type of execution - synchronous (sequential) or asynchronous (parallel) execution model.
Can be overridden by a value from input token, see the |
synchronous (default) | asynchronous |
Timeout (ms) |
no |
The maximal amount of time dedicated for a graph run; by default in milliseconds, but other time units may be used.
If the timeout interval elapses, the graph is aborted.
The value can be overridden by a value from an input token, see the The Timeout attribute is ignored for asynchronous execution type. Use the MonitorGraph component to watch the running graph. |
0 (unlimited) | positive number |
Input mapping |
no |
Input mapping defines how data from an incoming token overrides default execution settings. See Input mapping. |
CTL transformation |
Output mapping |
no |
Output mapping maps results of successful graphs to the first output port. See Output mapping. |
CTL transformation |
Error mapping |
no |
Error mapping maps results of unsuccessful graphs to the second output port. See Error mapping. |
CTL transformation |
Redirect error output |
no |
By default, results of failed graphs are sent to the second output port (error port). If this switch is set to |
false (default) | true |
Advanced |
|||
Auto-propagate graph parameters |
no |
Passes parameter values to executed graph (subgraph) without explicit mapping. If a parameter exists in both jobflow and executed graph and if the parameter is public in the executed graph, the parameter value is passed to the executed graph. |
no (default) | yes |
Execution group |
no |
The name of an execution group to which the executed graph belongs. An execution group can be used by KillGraph component as a named handler for a set of graphs to be interrupted. |
string |
Execution label |
no |
A text displayed in an execution view before a graph name. |
E.g. MyLabel |
Cluster node ID |
no |
A Cluster node ID which will be used for execution of a graph. |
string |
Execute graph as daemon |
no |
By default, all graphs are executed in non-daemon mode, so none of them can live longer than the parent graph. CloverDX Server automatically ensures for finished jobs that all non-daemon graphs are interrupted if they have not finished yet.
If you want to start a graph which can live longer than the parent graph, set this switch to |
false (default) | true |
Skip checkConfig |
no |
By default, the pre-execution configuration check of the graph is performed only if the check was performed on the parent job. This attribute can explicitly enable or disable the check. |
boolean (default is inherited from parent job) |
Stop processing on fail |
no |
By default, any failed graph causes the component to stop executing other graphs; information about skipped tokens is sent to the error output port. This behavior can be turned off by this attribute. Note: this function works only if an edge is connected to the component’s error port. |
true (default) | false |
Number of executors |
no |
By default, graphs executed in synchronous mode are triggered one after the other - the next graph is executed right after the previous one finishes.
This option allows to increment the number of simultaneously running graphs.
The |
positive number (1 is default) |
Locale |
no |
The default Locale of the executed graph. |
locale ID, see List of all Locale |
Time zone |
no |
The default Time zone of the executed graph. |
time zone ID, e.g. |
Details
The ExecuteGraph component runs a graph with specific settings, waits for the graph to finish and monitors the graph results and tracking information.
The .sgrf files can be run by the ExecuteGraph component as well as CloverDX graphs.
Execution Flow in Details
The component reads an input token, executes a graph based on incoming data values, waits for the graph to finish and sends results to a corresponding port. Results of a successful graph are sent to the first output port. Results of a failed graph are sent to the second output port (error port).
If the graph run was successful, the component continues with processing of the next input tokens.
Otherwise, the component stops executing other graphs, and from that moment on, all incoming tokens are ignored and information about ignored tokens is sent to the error output port.
The behavior related to processing after the first graph failure can be changed via the Stop processing on fail attribute.
If you set Stop processing on fail to false, the following graphs are executed.
Connected and disconnected ports
In case no input port is attached, only one graph is executed with default settings specified in the component’s attributes.
In case the first output port is not connected, the component just prints out the graph results to the log.
In case the second output port (error port) is not attached, the first failed graph causes interruption of the parent job. If you use Redirect error output, the disconnected error port does not cause interruption of the parent job.
Component configuration
For a graph execution, it is necessary to specify a graph location. Optionally, you can set up an execution type, initial values of graph parameters and dictionary content, timeout, execution group and several other attributes.
Most of the execution settings can be specified on the component via component attributes described bellow. These settings could be considered as a default execution settings.
The default execution settings can be changed dynamically and individually for each graph execution based on the data from the incoming token. The Input mapping attribute is the place where this override is defined.
|
Right-click the component and click Open Graph to access the graph that is executed. Similarly, in the ExecuteJobflow component, there is the Open Jobflow option. |
Execution type
The component supports synchronous (sequential) and asynchronous (parallel) graph execution.
-
synchronous execution mode (default) - the component blocks until the graph has finished processing, so graphs are monitored by this component until the end of run.
-
asynchronous execution mode - the component starts a graph and immediately sends the status information to the output. Graphs are only started by this component, the MonitorGraph component should be used for monitoring asynchronously executed graphs.
If the jobflow that executes the graph finishes earlier than the executed graph, the graph is killed. To avoid killing the graph, set Execute graph as daemon to true.
Input mapping
The Input mapping attribute allows to override the settings of the component based on the data from the incoming token. Moreover, initial dictionary content and graph parameters of an executed graph can be changed in Input mapping.
Input mapping is a regular CTL transformation which is executed before each graph execution. An input token, if any, is only input for this mapping. The transformation’s outputs consist of three records: RunConfig, JobParameters and Dictionary.
-
The RunConfig record represents execution settings. If a field of the record is not populated by this mapping, a default value from the respective attribute of the component is used instead.
Field Name Type Description jobURL
string
Overrides the component attribute Graph URL.
executionType
string
Overrides the component attribute Execution type.
timeout
long
Overrides the component attribute Timeout.
executionGroup
string
Overrides the component attribute Execution group.
executionLabel
string
Overrides the component attribute Execution label.
clusterNodeId
string
Overrides the component attribute Cluster node ID.
daemon
boolean
Overrides the component attribute Execute graph as daemon.
skipCheckConfig
boolean
Overrides the component attribute Skip checkConfig.
locale
string
Overrides the component attribute Locale.
timeZone
string
Overrides the component attribute Time zone.
jobParameters
map[string, string]
Graph parameters passed to the executed graph. Primary way of the definition of graph parameters is direct mapping to JobParameters record available in Input mapping dialog, where you can easily populate prepared graph parameters extracted from executed graph.
Graph parameters defined via this map have the highest priority. This map can be used in case the set of graph parameters is not available in design time of jobflow.
-
The JobParameters record represents all internal and external graph parameters of a triggered graph.
-
The Dictionary record represents input dictionary entries of a triggered graph.
|
JobParameters and Dictionary records are available in the transform dialog only if the component attribute Graph URL links to an existing graph which is used as a template for extraction of graph parameters and a dictionary structure. Only graph parameters and input dictionary entries from this graph can be populated by input mapping, no matter which graph will be actually executed in a runtime. |
Output mapping
Output mapping is a regular CTL transformation which is used to populate a token passed to the first output port. The mapping is executed for successful graphs. Up to four input data records are available for this mapping.
If the output mapping is empty, fields of RunStatus record are mapped to output by name.
-
The input token configuring the graph execution. This record is not available for a component without an input connector. This record has Port 0 displayed in the Type column.
This is very helpful for passing through some fields from input token to output token.
-
RunStatus record provides information about graph execution.
Field Name Type Description runId
long
A unique identifier of a graph run. In the case of an asynchronous execution type, the value can be used for graph monitoring or interruption.
originalJobURL
string
The path to an executed graph.
submitTime
long
Time when the graph execution was requested. The graph can be enqueued and started later, see
startTimebelow. See Job queue for more details about job queue.startTime
long
Time of graph startup. If the graph was enqueued, this is the time when the graph was taken from the queue and started. See Job queue for more details about job queue.
endTime
long
Time of graph finish, or null for asynchronous execution.
duration
long
Graph execution time in milliseconds.
executionGroup
string
The name of execution group to which the executed graph belongs.
executionLabel
string
The execution label with which the graph was executed.
status
string
The final graph execution status (FINISHED_OK | ERROR | ABORTED | TIMEOUT | RUNNING for asynchronous execution).
errException
string
Cause exception for failed graphs only.
errMessage
string
An error message for failed graphs only.
errComponent
string
The ID of the component which caused graph fail.
errComponentType
string
The type of the component which caused graph fail.
-
The dictionary record provides content of output dictionary entries of the graph. This record is available for the mapping only if the Graph URL attribute refers to a graph instance which has an output dictionary entry.
-
The tracking record provides tracking information usually available only in the JMX interface of the running graph. This record is available for the mapping only if the Graph URL attribute refers to a graph instance.
-
Tracking fields available for a whole graph:
Field Name Type Description startTime
long
Time of graph execution.
endTime
long
Time of the graph finish or null for the running graph.
executionTime
long
Graph execution time in milliseconds.
graphName
string
The name of an executed graph.
result
string
The graph execution status (FINISHED_OK | ERROR | ABORTED | TIMEOUT | RUNNING).
runningPhase
integer
An index of a running phase or null if the graph is already finished.
-
Tracking fields available for a graph phase:
Field Name Type Description startTime
date
Time of phase execution.
endTime
date
Time of the phase finish or null for running phase.
executionTime
long
Phase execution time in milliseconds.
memoryUtilization
long
Amount of used heap by the entire JVM at the time of phase execution (in bytes).
result
string
Phase execution status (FINISHED_OK | ERROR | ABORTED | RUNNING).
-
Tracking fields available for a component:
Field Name Type Description name
string
The name of the component.
usageCPU
number
Actual CPU time used by the component (user + system mode) expressed by a number from an interval (0, n) (0 means no CPU is used by the component, 1 means one CPU core is fully used by the component).
usageUser
number
Actual CPU time used by the component in a user mode expressed by a number from an interval (0, n) (0 means no CPU is used by the component, 1 means one CPU core is fully used by the component).
peakUsageCPU
number
Maximal CPU time used by the component (user + system mode) expressed by a number from an interval (0, n) (0 means no CPU is used by the component, 1 means one CPU core is fully used by the component).
peakUsageUser
number
Maximal CPU time used by the component in user mode expressed by a number from an interval (0, n) (0 means no CPU is used by the component, 1 means one CPU core is fully used by the component).
totalCPUTime
long
The number of milliseconds of CPU time used by this component (user + system mode).
totalUserTime
long
The number of milliseconds of CPU time in the user mode used by this component.
lifetimeMemoryAllocation
long
The cumulative number of bytes on heap allocated by this component during its lifetime.
result
string
The component execution status (FINISHED_OK | ERROR | ABORTED | RUNNING).
-
Tracking fields available for an input or output port:
Field Name Type Description byteFlow
integer
The number of bytes passed through this port per seconds.
bytePeak
integer
The maximal byteFlow registered on this port.
totalBytes
long
The number of bytes passed through this port.
recordFlow
integer
The number of records passed through this port per second.
recordPeak
integer
The maximal recordFlow registered on this port.
totalRecords
long
The total number of records passed through this port.
waitingRecords
integer
The number of records cached on the edge connected to the port.
averageWaitingRecords
integer
The average number of records cached on the edge connected to the port.
-
Error mapping
Error mapping is a regular CTL transformation. The fields are the same as in Output mapping.
Error mapping is used only if the graph finished unsuccessfully and the second output port is populated instead of the first one.
If Error mapping is empty, fields of the RunStatus record are mapped to output by name.
Examples
Executing graph
This example shows the basic usage of ExecuteGraph component.
Execute the graph invoice-processing.grf.
Solution
| Attribute | Value |
|---|---|
Graph URL |
${GRAPH_DIR}/invoice-processing.grf |
Best practices
|
If you drag a |
Compatibility
| Version | Compatibility Notice |
|---|---|
4.0.0-M2 |
You can now set up the Execution label attribute. |
4.1.0 |
You can now auto-propagate graph parameters. Empty output mapping and error mapping work as mapping fields by name:
|