Jump to: navigation, search

MonascaTrainPTG

Revision as of 13:30, 9 May 2019 by Witold.bedyk@est.fujitsu.com (talk | contribs) (Created page with "=== General === This page contains a summary of Monasca related subjects covered during the Train PTG. ;Place : Denver, Colorado ;Time : May 3-4, 2019 ;Etherpad : https://eth...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

General

This page contains a summary of Monasca related subjects covered during the Train PTG.

Place
Denver, Colorado
Time
May 3-4, 2019
Etherpad
https://etherpad.openstack.org/p/monasca-ptg-train
Priorisation game
https://docs.google.com/spreadsheets/d/1MWp8sCTeoJPnAysFzRKkxhvd8MR4AaJT04LxoJiE2QM/edit#gid=0


Development process

Actions
 * create a board for tracking Train tasks
 * clean up orphaned reviews and bugs


Threshold engine replacement

https://storyboard.openstack.org/#!/story/2005598

Option 1 - Prometheus remote read

https://docs.google.com/presentation/d/1tvllnWaridOG-t-qj9D2brddeQXsYNyZwoYUfby_3Ns/edit?usp=sharing

Pros
replaces thresh, notification and transform components
relatively low effort
Cons
creates strong dependency on Prometheus
not clear how to distribute configuration to instances (no API available)
not clear about performance and resources usage of remote read

Option 2 - New stream processing application

https://faust.readthedocs.io/

Pros
keeps existing architecture
scalable, HA application implemented in Python
only Kafka dependency (no Storm)
new implementation would integrate the new QL
Cons
large effort
Actions
 * evaluate plausibility of Faust based implementation
 * ask decision makers for developer commitments


Reuse Prometheus dashboards

https://storyboard.openstack.org/#!/story/2005617

Pros
leverage community resources
low effort
Cons
additional (but optional) Prometheus dependency


New Query Language

https://storyboard.openstack.org/#!/story/2005618

Pros
Extends querying capabilities
Makes transform (aggregation) engine obsolete
Prometheus dashboards can be easily adapted
Cons
Cassandra support probably expensive
Actions
 * create spec
 * ask decision makers for developer commitments


High availability/scalability for InfluxDB

https://storyboard.openstack.org/#!/story/2005620

  • make use of Kafka to replicate and partition the measurements
Alternatives
existing Apache Cassandra
add Gnocchi support
Actions
 * Document HA options on d.o.o


Application Credentials

https://storyboard.openstack.org/#!/story/2005622
https://storyboard.openstack.org/#!/story/2005623

  • simplify secure tenant agent deployment
  • decouple Grafana datasource from Grafana auth
Actions
 * Add scripts for Grafana configuration to monasca-grafana-datasource


Grafana Loki

Actions
 * Investigate as possible alternative for ELK stack


Vitrage integration

https://storyboard.openstack.org/#!/story/2004550 https://etherpad.openstack.org/p/vitrage-train-ptg

Agreement
* Implement alarm based mapping to Vitrage resources
Actions
 * Update spec
 * Merge initial patch


Middleware upgrades

https://storyboard.openstack.org/#!/story/2005624

  • Apache Kafka
  • InfluxDB
  • Apache Storm
  • ELK


Openstack Helm

Actions
 * Aim Monasca lite deployment
 * Create story