why monitoring sucks and what configuration management can do about it
TRANSCRIPT
Why Monitoring SucksAnd what Configuration Management can do about it
James Turnbull@kartar
1
Who• VP of Engineering at Kickstarter
• Advisor at Docker
• Open source chap
• Funny accent
2
Who are you folks?
5
What's this all about?
6
Monitoring Sucks
7
It fucking does and I can prove it.
8
Data. We have data.
9
Step out of the bubble
10
The wrong folks monitor
11
Only 9% of developers do
monitoring
12
We monitor the wrong things.
13
68% of folks don't monitor business
logic
14
For the wrong reasons
15
90% have unmonitored failures
16
80% have un-actioned alerts
17
33% monitor reactively
18
Our tools don't scale
19
They're centralized
20
And not self-service
21
So why?
22
Well ... lots of "maybe"
23
Horses hate talking about unicorns
24
A few unicorns do not make horses magic
25
Change is slow
26
ROI isn't as obviousInvestment is expensive
Don't know what you don't know
27
But there is hope
28
Configuration Management as
default
29
73% of y'all use it
30
Aren't y'all awesome?
31
How can we take advantage of this?
32
Blessed are the tool makers
33
Events
34
APIs and endpoints
35
Create monitoring abstractions
36
Build monitoring patterns into content
37
Set the quality bar to include monitoring
38
Testing compliments monitoring
39
Make service discovery and
registration a first class citizen
40
Build tools that enable self-service
41
Questions?
42