Monitoring the eG Manager
The eG Enterprise provides a dedicated eG Manager monitoring model that monitors each of the core functions of the eG manager at configured frequencies and reports abnormalities (if any).
Each layer of Figure 1 above is mapped to tests that report a wealth of performance information related to the eG manager. To pull out these performance metrics, the eG agent can be deployed on the eG manager host itself (agent-based) or on any remote Windows host in the environment (agentless).
Regardless of where it is installed, the eG agent should be configured to connect to the JRE used by the eG manager to pull out metrics of interest. To enable the eG agent to connect to the JRE, by default, JMX support has been enabled for the eG manager’s JRE. JMX enables external programs like the eG agent to connect to the JRE of an application and pull out metrics in real-time.
With the help of the metrics collected via JMX, eG administrators can find quick and accurate answers for the following queries:
- Has database cleanup run today?
- Did cleanup take too long to complete?
- Was cleanup incomplete on any table? If so,which ones?
- Did cleanup fail on any table? If so, which ones?
- Does the database connection pool have enough free connections?
- Are too many connections to the database waiting to close?
- Does the eG database have too many large sized tables and tables with many rows?
- Are any eG agents not running currently?
- Have any eG agents configured on the manager not being physically installed?
- Did any errors occur on the eG manager? If so, when - during manager installation/configuration/starting? during threshold computation? or at the time of trend computation?
- Have all object pools on the eG manager been correctly sized? Are there any object pools that are running out of free objects?
- Is any servlet responding poorly to client requests?
- How busy is the state manager process?
- How is the threshold manager performing? Is it taking too long to compute thresholds?
- On what tables did threshold computation fail and why?
- Did trend run today?
- Is the trend manager process healthy? Is it taking an unusually long time to compute trends?
- Did trend computation fail on any table? If so, which ones?
- Are cleanup/threshold computation/trend computation performed by separate Java processes or by the eG manager itself?
- How many users are registered with the eG Enterprise system and how many user subscriptions are about to expire?
- Is the eG manager overloaded with sessions? If so, what are the sessions that are currently active on the manager?
- Who are the distinct users who are logged into the manager, and when did they login?