[FINERACT-880] Fineract Caches & Kubernetes Deployment

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: Performance
• Labels: kubernetes scalability technical
• Assignee: Benura Abeywardena
• Reporter: Michael Vorburger
• Created: 12/Apr/20 09:45
• Updated: 22/Jan/21 05:00

Description

The Kubernetes deployment contributed in FINERACT-783 by default creates a Kubernetes Deployment with a single Pod (based on Fineract's Docker Compose set-up).

The goal of this issue is to investigate what is missing to make it possible to "scale horizontally" by increasing the number of replicas.

For example, I believe Fineract code caches certain things? I'm not clear if they are enabled or disabled by default. Such caches typically do not "just work" if you scale the Kubernetes deployment, because they will cause "cache inconsistencies", unless the caches collaborate to invalidate entries when appropriate. This probably need some configuration.  Some references:

• Log about some EhCache configuration when Fineract starts
• cache.xml and ehcache.xml in Fineract source code
• package org.apache.fineract.infrastructure.cache in Fineract source code
• kumaranath_fernando (kumaranathfernando ?)'s GSOC 2018 proposal and final report
• apremalal's GSOC 2013 work about caching: YouTube demo, Mifos.org blog,
• "Data Layer" on https://cwiki.apache.org/confluence/display/FINERACT/Key+Design+Principles

I personally do not know the details about caching in Fineract. We should either fine existing create new documentation about this.  It would be worth to reach out to the mailing list to start a discussion and get input on what is required to make this work. 

FYI xurror, awasum, angeh

Issue Links

• is blocked by: FINERACT-783 Fineract on Kubernetes
• is blocked by: FINERACT-830 Use distroless base image instead of bitnami/tomcat in container
• is blocked by: FINERACT-135 Performance improvement using EHCache or any other cache managers