Melvyn Sopacua & Folkje Lips - MM15NL

Post on 13-Apr-2017

259 views 3 download

Transcript of Melvyn Sopacua & Folkje Lips - MM15NL

Practical New Relic:

Save hours hunting

millisecondsFolkje Lips & Melvyn Sopacua

Melvyn Sopacua

Senior Support Engineer

@ SupportDesk

&

Folkje Lips

Product Owner Hypernode

@ Byte

Hypernode

Speed up your shop with Hypernode

Speed up even more with New Relic

Performance monitor

Error monitor

Uptime Monitor

Alert Dispatcher

Cannot predict revenue :(

Look and Feel

Transaction overview

Breakdown table

Three main uses for devs and maintainers

1)Performance optimization track (project based)

2)Bug or performance bottleneck analysis (incidents)

3)Performance monitoring over time for scalability

Performance Optimization Track

Recursive loads > Expensive observers > Platform bottlenecks > Too Much Bling > Shaving the milliseconds

Recursive Load

Many Recursive Loads

Many Recursive Loads

Queries

Basic Workflow

●Identify routerpaths that are worth investigating (slow versus time consumed)

●Create a Key Transaction and set it's desired load time to customer's goal or four times less.

●Wait for transaction traces to come in and analyse bottlenecks.

●Buy a new template Fix the code

●Check gain when live or watch the new bottleneck rear it's ugly head.

Identifying Routerpaths

Bug or performance bottleneck incident

This guy knows why.

The cannot reproduce problem

●“What did you do to break it?”

●“Works for me, so call back when you see it again”

●“Site's up and running” (and you forgot your /etc/hosts file)

●7 day error retention

●Graphed

●Alertable (based on % of requests)

●Traces

Error rate screen

Error trace

Performance bottleneck

Performance monitoring

Alerts, SLA's and Active Checks

Performance Monitoring Toolbox

Event alerts Periodics

Application wide ApDex

Key Transaction ApDex

Error rates

Reachability

Weekly Performance Email

90 days application wide

30 days per transaction

Historic Performance Comparison

Alert System

●Schematic alert policies goes here●Schematic alert policies goes here (it's being changed)

Reporting the good stuff

Reporting the bad stuff

●Still looking for good example

Service Level Agreements

●Is changing with alerts, have to redo

Interested?

●Price directly at New Relic Price for same Pro at Byte