Write up "Stacks"

[mbbroberg]

"The Stacks"

There are a few maturing visions of a complete solution that goes from Telemetry all the way to Alerting.

Stack Name	Sensors	Metrics	Logging	Telemetry	Persistence	Querying	Visualization	Notification	Alerting
Raintank	plugins	Snap	:x:	Graphite	Grafana	Grafana	Grafana	Grafana	Grafana
ELK	plugins	Beats	Logstash	Elasticsearch	Elasticsearch	Kabana
Prometheus	clients
Nagios	plugins	Nagios			:x:	:x:	Nagios	Nagios	Nagios
Zabbix
Netflix
VMware

Open Source? Enterprise Support?

Use information from this presentation A Working Theory of Monitoring:

Nagios

Sensing:
Subprocesses and plugins, LOTS of plugins
Collection:
Centralized scraping
Support for forwarding metrics
Analysis: At sensing time
Alerting:
Configurable alarms and emails
Visualization:
Basic graphs of check results
Dependency chains

Ganglia

Sensing:
gmond on nodes
extensions/plugins
Collection:
multicast, UDP, TCP polls
Analysis:
value_threshold
external (nagios)
Storage: rrdtool/rrdcached
Alerting: N/A
Visualization: ganglia-web

Sensu

Sensing: Arbitrary JSON emitters “Checkers”
Collection: RabbitMQ JSON event bus
Analysis:
Handlers
Storage: N/A
Alerting:
Handlers
Visualization: N/A

ELK

Sensing:
Deployable log thrower
Collection:
MQ (Redis)
Analysis:
Indexer
Storage:
ElasticSearch
Alerting: N/A
Visualization:
Kibana (ES)

OpenTSDB

Sensing:
Custom clients
Collection:
TSD RPC
Analysis:
External
Storage:
Complete storage layer
Alerting: N/A
Visualization: N/A

Graphite

Sensing:
DIY, name+value
Collection:
Custom messaging protocol
Analysis: N/A
Storage: Carbon+Whisper
file-per-metric
Alerting: N/A
Visualization:
Static config of complex graphs

[mbbroberg]

Part of this is discussing how we go beyond monitoring...

Going Beyond Telemetry

Sensors -> Telemetry -> Persistence -> Learning -> Visualization -> Notification -> Alerting