[IGNITE-6814] Detailed memory consumption on start and OOM reporting

• Type: Improvement
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 2.0
• Fix Version/s: 2.4
• Labels: iep-6 important usability
• Assignee: Vyacheslav Koptilin
• Reporter: Denis A. Magda
• Created: 01/Nov/17 23:08
• Updated: 09/Jan/18 17:35
• Resolved: 09/Jan/18 17:35

Description

Presently Ignite allocates 20% of RAM on a node startup, however, the user doesn't see automatically chosen memory settings. Also, if there a node runs out of RAM and throws an OOM error there are no hints on why this happened and how to fix it.

Suggestions:

• Add off-heap field to Topology Snapshot message. The field will accumulate maximum size of all the data regions defined cluster-wide:

  Topology snapshot [ver=1, servers=1, clients=0, CPUs=4, off-heap={N}, heap=1.8GB, ]
  

• Print detailed memory configuration below Topology Snapshot message following this format:

  Data Regions Configured:
  ^-- Data_Region_Name [initSize=N, maxSize=N, persistenceEnabled={true|false}]
  

  Example:

  Data Regions Configured:
     ^-- Default [initSize=100MB, maxSize=5.0 GB, persistenceEnabled=true]
     ^-- RegionalMetrics [initSize=500MB, maxSize=15.0 GB, persistenceEnabled=false]
  

• Provide guidelines on how to overcome OOM when it happens. Specify a data region name with all its current parameters and suggest three possible things - tweak maximum memory, setup eviction policies or enable Ignite persistence.

  Out of memory in data region Region_Name [initSize=N, maxSize=N, persistenceEnabled={true|false}]. Do one of the following:
     ^-- Increase maximum size (DataRegionConfiguration.maxSize)
     ^-- Enable Ignite persistence (DataRegionConfiguration.persistenceEnabled) or
     ^-- Enable eviction or expiration policies.
  

Issue Links

• is part of: IGNITE-5796 Improve cache memory metrics calculation and activation
• links to: pull-request
• links to: Review