Discuss best practices for obtaining baseline Librato statistics

[elopatin-uc3]

IAS is asking that each program performs quarterly checks of systems via Librato.

Capture baseline statistics up front so we have data to base future evaluations on for each microservice.

Recap on past incidents; noting what caught our eye during each one so we know what we would want to pay attention to for specific microservices.

Team should establish its timeline for doing this on a quarterly basis; start in 1-2 months.

Librato: based on SAR

Ashley is involved in IAS meetings; sync up on her use of metrics on Merritt systems in recent past; suggestions going forward.

[marisastrong]

https://confluence.ucop.edu/display/CUG/Periodic+AWS+Instance+Reviews

[marisastrong]

CloudMetrics drive the Librato views Could pull data directly from CloudMetrics API to munge/massage the data Can also pull application statistics into CloudMetrics that can be viewed in Librato EZID is doing this Memory is not provided by default - IAS did a customization to include this sar metric at the OS level

[terrywbrady]

Librato is a visualization of CW metrics. It requires you to provide counts/summaries. It does not really compute counts.

• Recommendation: have apps compute metrics to 5 minute windows.
• There is a premium charged to get metrics below 5 minutes.

OpenSearch will also pull from CW logs.