1
00:00:00,167 --> 00:00:04,000
At two in the morning, an automated tool lowers a service's

2
00:00:03,900 --> 00:00:05,700
memory limit in production.

3
00:00:05,600 --> 00:00:08,267
By six, the service is restarting.

4
00:00:08,167 --> 00:00:10,100
Who approved that change?

5
00:00:10,567 --> 00:00:13,700
Automatic changes act on a model of your workload.

6
00:00:13,600 --> 00:00:16,267
They do not know about next week's launch,

7
00:00:16,167 --> 00:00:20,133
a migration in progress, or a promise made to a customer.

8
00:00:20,600 --> 00:00:22,500
Evidence should inform a change.

9
00:00:22,400 --> 00:00:25,067
Engineering judgment should decide it.

10
00:00:25,533 --> 00:00:29,300
In RATIVERA, every recommendation waits for a person:

11
00:00:29,200 --> 00:00:31,067
accept, reject, or defer.

12
00:00:30,967 --> 00:00:35,300
Each decision is recorded against the assessment it was based

13
00:00:35,200 --> 00:00:35,900
on.

14
00:00:35,800 --> 00:00:38,233
Accepting records the decision only.

15
00:00:38,133 --> 00:00:42,000
It does not execute or authorize an infrastructure change,

16
00:00:41,900 --> 00:00:45,133
and RATIVERA does not modify your cluster.

17
00:00:45,600 --> 00:00:48,667
The change itself belongs in your Git workflow:

18
00:00:48,567 --> 00:00:51,067
a pull request that your team reviews,

19
00:00:50,967 --> 00:00:54,300
merges, and deploys through its own pipeline.

20
00:00:54,200 --> 00:00:57,067
That way, every production change stays traceable:

21
00:00:56,967 --> 00:01:00,467
who proposed it, who approved it, and when it shipped.

22
00:01:00,933 --> 00:01:05,333
RATIVERA. Understanding Kubernetes, one problem at a time.