Performance Analytics score collection follows the process described here. To aid
troubleshooting, a mapping between job steps and log entries is provided.
Collection overview
The data collection job involves executing an SQL query for each indicator source that uses
the data collector. The query repeats for every collection time from the start date to the
stop date. Then queries run for the next indicator source. Each step of executing a query is
documented in the data collection job log.
Figure 1. High-level view of data collection job
Optimization properties shown in log
The data collector has several optimizations that can be turned on or off in the System
Properties. The beginning of the job log includes the state of these properties.
Note:
Do not confuse these optimization properties with the new, optimized data collector.
These properties apply only to the classic data collector. The new data collector is optimized
at a more fundamental level.
Table 1. Data collection optimization properties
Optimization
Description
System property
Nested collection
Replaces flat mapping, to reduce redundant objects. For example, instead of creating
identical indicator objects for every breakdown and element on that indicator, only one
object is created for the indicator.
com.snc.pa.dc.nested_collection
Default: true
Flush snapshots
The ServiceNow AI Platform stores a maximum number of Performance Analyticssnapshots in memory before offloading them to the database.
com.snc.pa.dc.flush_snapshot_count
Default: 1000
Use hashcode for count distinct
Stores count distinct aggregations as a hashcode instead of as separate
values.
com.snc.pa.dc.use_hashcode_for_distinct
Default:
true
Starting and completing collection job
Follow the data collection process as revealed in log messages. The log messages in each
subprocess have been separated into their own tables.
Note:
This log is for the original data collector. If your instance uses this data collector, the
system parameter com.snc.pa.dc.hsql exists and is "false." Otherwise, you
are using the optimized data collector as described in Optimizing data collection. For more information about that log, see
Log details for optimized data collector.
Collection job step
Example of resulting log entry
Shows the domain of the collection job Run As user.
User <admin> belongs to domain global
Collection starts. Job start specifies whether scores, text index, or both are being
collected. Messages also specify the settings of the data collection optimization
properties.
Starting collection of Both scores and text index. Value of properties in system
-
Flush Snapshot Count: 1,000
Is nested collection: true
Using hashcode for
count distinct: true
Shows the domains for which scores are collected. Also shows other relevant details
about the domains.
Lists each indicator associated with the job and the indicator properties that are
related to data collection:
Are records collected in a snapshot?
Are any breakdowns excluded?
Are the collection periods in the indicator source overridden?
Is the breakdown matrix collected?
If the breakdown matrix is collected, are any breakdown combinations excluded?
JobIndicator: Number of open incidents - de537dc3d7131100b96d45a3ce610305 Job
Indicator Properties: collectIndicator?: Yes, BreakdownCollectionType: WITH_EXCLUDE Excluded
Breakdowns: Assignment Group, State Indicator Properties => CollectingRecords: Yes,
Overridden Collection Periods: No, Scripted?: No, Breakdown Combinations?: Yes, Excluded
Breakdown Combinations: No combination is excluded.
List indicators that have additional conditions, beyond the conditions inherited from
the indicator source.
Indicator: Number of reassigned open incidents - ffb59561ff2302001e68ffffffffff33,
Additional Conditions: reassignment_count>0
Indicates whether score collection is following the nested data structure instead of
using flat mapping. Nested collection creates fewer redundant objects in
memory.
Non-nested collection supports batching of scores.
Using nested collection.
Note - Batch inserts of scores are not supported with
nested collection.
Collecting for 20191103 on Indicator Source Incidents.Open
Fetch the following fields:
The unique fields of the breakdown sources (typically sys_id)
The indicator fields used in breakdown mappings
The table fields used in scripted breakdown mappings
The indicator source fields used in text analytics, if the text index is collected
(typically includes short_description)
Fetching "short_description,sys_id, opened_at, assignment_group, description,
priority, category" from "incident"
Generate SQL based on the conditions that are specified in the indicator
source.
Note:
If the indicator source specifies Today in one of the conditions, Today is
considered relative to the period for which the data collection job is executed. For
example, the Incidents.New indicator source includes the condition
[Opened][on][Today]. With days defined to start at 07:00:00, when
data is collected for 2019-10-28, the job produces the SQL script on the
right.
SELECT task0.`sys_id` FROM task task0 WHERE task0.`sys_class_name` = 'incident' AND
(task0.`opened_at` >= '2019-10-28 07:00:00' AND task0.`opened_at` <= '2019-10-29
06:59:59')
Display the number of records fetched by the SQL.
Fetched 150 rows from Indicator Source Incidents.Open (facts table: incident)
Run the map/reduce function to calculate indicator scores.
If text indexing is active and has been configured for the indicator
source, the data collector stores the resulting text index.
Storing Text Index for indicator source Incidents.Open
Bytes used by text index: 41,984 for: Incidents.Open
Begin storing newly collected results for the indicator source.
Scores to be stored: 374,842 (excluding lazily evaluated level 2 count distinct scores
for indicators - ID - All Incidents with All 3 Breakdowns)
Any indicator with a null value is logged.
Stored 'Value when nil': 0.0 as score for indicator: Number of incidents deflected by
knowledge
Finish storing results.
Scores stored: 1,125,057
Stored collected results
Specify which indicators the data collector does not collect scores for.
Not collecting for Indicator: Summed age of open incidents with excluded Breakdown:
Assignment Group
Finish collecting data for that indicator source for that period.
Collection for 20200519 on Indicator Source Incidents.Open finished
Display the average amount of time spent on a record to process additional conditions
that were set on indicators.
Average time taken to process additional indicator condition for each record: 138
ms
Steps of the map/reduce function
Collection job step
Example of resulting log entry
The map/reduce function runs.
Applying map/reduce function for indicator source Incidents.Open
Warning if text indexing is active for the job but cannot run because of a missing or
invalid text index configuration.
Skipping text index collection as there is missing or invalid text index
configuration
Delete previous scores for the indicators and breakdowns that use the
indicator source.
Deleting previous results for indicator source Incidents.Open
Deleted previous results 38 for indicator: External Logins
Deleted previous results 21 for indicator: Failed Logins
Process each record on the indicator source to calculate scores. Log
messages track running total of processed records and memory usage. Processing for each
record involves:
Execute or evaluate any scripts.
Validate indicator conditions.
Calculate the indicator score.
For each breakdown:
Calculate the breakdown score or execute the breakdown script.
Retrieve all breakdown unique values.
Create or update the array for scores or snapshots. The array is: Indicator,
Breakdown 1, Artifact 1, Breakdown 2, Artifact 2, Domain, Value, Array of
[sys_id]
Processed 18%% (1) records, Result Object size: 5 MB
Processed 42% (2) records, Result Object size: 9 MB
Processed 76% (3) records, Result Object size: 14 MB
Processed 100% (4) records, Result Object size: 20 MB
