How to View the KPI Dashboard

You can use the key performance indicator (KPI) dashboard in the JMX console to view the following information of the UCMDB Server:

  • Database connection summary
  • Discovery processing statistics
  • Operations information

Follow these steps to view the KPI dashboard:

  1. Launch a web browser and enter the following address: https://localhost:8443/jmx-console.

    You may have to log in with a user name and password.

  2. Locate KPI:service=Dashboard and click the link to open the JMX MBEAN View page.

  3. Invoke one of the following methods:

    • viewDbConnectionSummary: displays database connection summary
    • viewDiscoveryProcessingStatistics: displays discovery processing statistics
    • viewOperationsInformation: displays operations information

    Note The showHTML option controls if the information is displayed in the HTML or JSON format.

    True: the information is displayed in the HTML format

    False: the information is displayed in the JSON format

If the database is Oracle, the viewDbConnectionSummary method can gather and display database sessions. To enable this capability, the customer DBA must grant rights to the UCMDB:

  • grant select on v_$session to <db_schema>;
  • grant select on v_$locked_object to <db_schema>;
  • grant select on v_$sqlarea to <db_schema>;
  • grant select on v_$transaction to <db_schema>;

The UCMDB Server regularly exports the information displayed in the KPI dashboard into the following folder: UCMDBServer/runtime/log/statistics/metrics_json. The automatic export is performed by a scheduled job: KPI Dashboard Dump, which by default runs hourly. The exported files have names that resemble the following examples:

  • DbConnectionSummary2016.05.01-11.30.00.json
  • DiscoveryProcessingStats2016.05.01-11.30.00.json

If you want to start or stop this automatic export, use one of the following options:

  • Activate or deactivate the KPI Dashboard Dump job from Administration > Scheduler
  • Set the statistics.gathering.enabled setting to true (activate) or false (deactivate).

Performance Statistics Dashboard

The UCMDB Server automatically captures performance metrics and saves the data statistically in .csv files under the UCMDBServer\runtime\log\statistics\metrics folder.

To define the interval (measured in minutes) at which the UCMDB Server gathers performance metrics:

  1. In UCMDB, go to Administration > infrastructure Settings Manager > General Settings.

  2. Locate Statistics gathering interval setting and specify a time interval in minutes. The default value is 15 minutes.

  3. Restart the UCMDB Server.

The following table explains each column in the .csv files.

Column Information
t Timestamp (epoch)
count Number of invocations 1
max Max invocation duration 1
mean Average invocation duration 1
min Min invocation duration 1
stddev Standard Deviation of the values
p50 50% quantile 12
p75 75% quantile 12
p95 95% quantile 12
p98 98% quantile 12
p99 99% quantile 12
p999 99.9% quantile 12
mean_rate Average event rate (per second)
m1_rate Average event rate in the last minutes
m5_rate Average event rate in the last 5 minutes
m15_rate Average event rate in the last 15 minutes
rate_unit Unit for the rate (for example, calls/second)
duration_unit Unit for the durations (for example, seconds)

Note  

  1. All these values are collected since the last metric reset. Every 15 minutes after dumping the statistics to a CSV, the metrics (counters, durations) are reset to 0.
  2. A quantile is calculated by taking the entire data set, sorting it, and taking the value in the middle (or 1% from the end, for the 99th percentile). For example, the 75% quantile shows the median of the lower 75% values. These quantiles apply to the set of durations of each invocation.

Counters are not reset every 15 minutes, so their values keep growing and show overall counts since the system startup.

ClosedCaptured Metrics

The following tables show the captured metrics.

Data-in statistics
Metrics

Measurement

 Note
data-in.<submodule> Timer A breakdown of times in the data in process, by submodule (and submodules of submodules)
data-in.total Timer An overall timer from start to end for the data-in operation

 

UCMDB API Request statistics
Metrics

Measurement

 Note
api-executed.<Operation> Timer A timer over API requests
api-failed Counter A counter that shows the number of failed API requests

 

Server Session statistics
Metrics

Measurement
server-sessions Counter

 

DB Connection statistics
Metrics

Measurement

db-connection-borrowed

 Counter

db-connection-borrow-failed

 Counter

db-connection-returned

 Counter

db-connection-invalidated

 Counter
Connection timers Timer

 

DAL statistics
Metrics

Measurement

 Note
dal-executed.<sql> Timer A timer over the execution of a specific, long-running SQL (>1 minute)
dal-executed Timer Overall timer across all DAL executions
dal-failed Counter  
db-rows-updated Counter A counter showing how many rows all the SQLs modified

 

TQL statistics
Metrics

Measurement

 Note
tql-calculation.<tql> Timer A timer over the execution of a specific, long-running TQL (>1 minute)
tql-calculation Timer Overall timer across all TQL executions
tql-results Timer Counter for the total amount of CIs returned by TQL queries so far

 

Startup statistics
Metrics

Measurement

 Note
startup Timer A timer over the startup of UCMDB and shows the time needed for the server startup

 

Operation statistics
Metrics

Measurement

 Note
operation-executed.<"sync"/"async">. <CallerApp>.<OperationName> Timer

Timers over specific operations

  • “sync”: the caller is waiting for the operation to finish
  • “async”: the operation runs in an async thread
operation-executed."end-to-end" .<"sync"/"async">. <CallerApp>.<OperationName> Timer Timers that are over end-to-end operations, meaning the operation in the timer name is the first operation in an operation call stack
operation-failed Counter A counter that shows the number of failed operations

 

UI Server statistics
Metrics

Measurement

 Note
ui-service-executed.<Service> Timer Timers over specific UI Server service calls
ui-service-failed.<Service> Counter  

 

JVM statistics
Metrics
"gc"
"buffers"
"memory"
"threads"

 

Jetty statistics
Metrics
war-context-handler.<context>
root-handler
empty-root-handler
jetty-connections

 

Jetty Queued Thread Pool
Metrics
utilization
utilization-max
size
jobs

 

Statistics on overall counters (objects, links, merges) for data-in operations
Metrics Measurement
merged-cis Counter
type_changes Counter
merge-operations Counter
ignored-from-cmdb Counter
ignored-from-bulk Counter
updated-objects Counter
updated-links Counter