Description
Information captured
Create a context tracer on stall
Create a context tracer on every call
Configure the data collected
Configure the data displayed in the portal
Screenshot

Description

To help troubleshooting analysis, HPA provides detailed information context named Context Tracer in two cases:

on some specific request executions, configured by a limit of execution time (stall), in agent configuration
for all executions of a specific request, configured with a specific custom metric declaration using the dedicated HpaDumpParamTracer
- but no new metric captured

Information captured

In both cases, HPA captures and records:

the object that execute the method metric
the parameters of the method metric
the return value of the method metric
the HPA metric(s) involved

When a context tracer execution comes from a stall, then HPA captures also:

the stacktrace
the thread information
start/end/elapsed time of the execution
calypso session information when available
the stall indicator: is the method execution time is over the configurable threshold

see the screenshots for an overview

Create a context tracer

Context tracer on stall records a stall metric AND a context tracer.

Context tracer on stall

To record a Context Tracer on a long execution time, configure the agent with

override_config/configuration_stall.yml

This file describes the metrics to record:

either by giving the metric name
or by using a regular expression

Then, in the same block:

set a execution time limit
set the action to do:
- NOTHING to ignore the stall (can be useful to clean the data recorded)
- METRIC to capture the stall metric only
- METRIC_AND_CONTEXT to capture both the stall metric and the related context tracer
set the metric(s) filter
- DIRECT_VALUE to use the full metric name
- REGEXP to select multiple metrics at once

The combination action-stall duration-filter defines a perimeter for which the HPA expected behavior is defined.

There is also a default action with a default time execution limit.

actions-on-stall:
  default-action:
    action: METRIC_AND_CONTEXT
    stall-duration-in-seconds: 30
  custom-actions:
    - action: NOTHING
      stall-duration-in-seconds: 60
      filter: DIRECT_VALUE
      metrics:
        - Frontends|Apps|TimerThread
    - action: METRIC
      stall-duration-in-seconds: 180
      filter: REGEXP
      metrics:
        - .*PThread.*

Context tracer on every call

This use case is used to record all the calls made on a method, for example:

record all bad events on engines
record all report parameters on navigators

Follow the steps to enable it :

create a custom metric
with the following definition :

#
# Context tracer on <com.calypsox....mypackage.MyClass>.<myMethod>
#
SetFlag: CustomerCTNewFlag
TurnOn: CustomerCTNewFlag

IdentifyClassAs: <com.calypsox....mypackage.MyClass> CustomerCTNewFlag

TraceOneMethodWithParametersIfFlagged: CustomerCTNewFlag <myMethod> HpaDumpParamTracer "CustomerCTNewFlag"

The class <com.calypsox....mypackage.MyClass> is monitored on the method <myMethod>. Each time the method is executed the context tracer will record the usual information.

Configure collected data

When the context tracer is triggered, it records many objects information.

The way those objects information are captured is described in default_config/configuration_context_tracer.yml

To avoid/limit the overhead, there are 2 available configuration

a limit of the number of context tracer recorded by the agent: maximum-trace-sent-every-15s
a limit when HPA detects collections : maximum-elements-to-get-in-collections
- When a parameter or return value is a list, it records only X elements in the list.

There is already a list of objects that are recorded in details, to modify this list and/or the above limites, configure the agent in

override_config/configuration_context_tracer.yml

It overrides the default file, so keep the actual configuration if needed in the override file.

maximum-trace-sent-every-15s: 50
maximum-elements-to-get-in-collections: 2
conversions:
  "com.calypso.tk.bo.BOCre":
    - property: "getId"
      converter: METHOD
    - property: "getStatus"
      converter: METHOD
    - property: "getCreType"
      converter: METHOD
  "com.calypso.tk.util.InventorySecurityPositionArray":
    - property: "firstElement"
      converter: METHOD
    - property: "lastElement"
      converter: METHOD
    - property: "size"
      converter: METHOD
  ...
  ...

Configure displayed data

Context tracer are stored in a JSON structure in the database, to ease the analysis in the portal, it possible to extract and display some values in columns.

Custom columns

There is already a list of objects that are displayed in details, to modify this list, configure on the Portal :

override_config/traces/trace-configuration.yml

There is up to 15 custom columns with :

id: column1, column2 …. to column15
name: the name of the column
config: the data to display from the JSON
- type: how the data is searched
  - ANYLEVEL : in the JSON, find the property identified by the text in value
  - REGEXP : in the JSON, find the properties identified by the regular expression in value

Here is an example of the file :

configurationTraces:
- id: column1
  name: 'Trade id'
  config:
	- type: ANYLEVEL
	  display-name: 'trade_id'
	  value: com.calypso.tk.core.Trade.getLongId
	- type: ANYLEVEL
	  display-name: 'trade_id'
	  value: com.calypso.tk.event.PSEventTrade.getObjectId
- id: column2
  name: 'Trade status'
  config:
	- type: ANYLEVEL
	  value: com.calypso.tk.core.Status.toString
- id: column3
  name: 'product_id'
  config:
	- type: ANYLEVEL
	  value: com.calypso.tk.core.Product.getId
	- type: REGEXP
	  value: 'com.calypso.tk.product.*.getId'
- id: column4
  name: TradeFilter
  config:
	- type: ANYLEVEL
	  value: com.calypso.tk.mo.TradeFilter.getName
- id: column5
  name: 'Book name'
  config:
	- type: ANYLEVEL
	  value: com.calypso.tk.core.Book.getName

Screenshots

Full page - context tracer

Event :

Zoom PSEvent - context tracer

HPA Portal sections Context tracer