1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
|
I'm more and more annoyed by how the tech community is super
enthusiastic about new pieces of technology, and how hard they try to
convince you it's the best next thing in the world. Way too often, at
conferences or meet-ups, the talks tend to glorify a product or a
technology, and only focus on how it will make your life easier. It's
too common to have someone do a demo on stage on how to build, in 5
minutes, a trivial application running with X many instances in a
container in the cloud and be like "see how easy it was !?".
What will not be mentioned is how your team is going to transition to
this technology or infrastructure; what are the traps hiding; what will
not work; what are the real limitations (can it scale to more than 10
instances ? 100 instances ? 10k instances ?); how do you manage it in
your data-center; in your cloud; how easy is it to debug; what are the
current issues that people running it in production have met; what's the
worst case scenario for an incident; how long can it take to recover;
and way too many other things.
Over the last few days, I binge-watched many of the
[[https://www.youtube.com/playlist?list=PLVjgeV_avap2arug3vIz8c6l72rvh9poV][MesosCon]]'s'
videos. One of the talk I really enjoyed was by
[[https://twitter.com/Yasumoto][Joseph Smith]]. In
[[https://www.youtube.com/watch?v=nNrh-gdu9m4&index=8&list=PLVjgeV_avap2arug3vIz8c6l72rvh9poV][his
talk]], he shared about various ways Mesos and Aurora failed at Twitter.
Joseph's talk was the opposite of what I described earlier. He mentioned
at length issues and problems we've encountered running Aurora. Some of
the issues he explored were recent (from a couple of weeks ago); some
were pretty old and are fixed by now; and also what would be the worst
case scenario that could happen. This is exactly what I want to hear
when someone introduces a piece of technology. I need to be aware of
them. It doesn't mean that I'm going to be scared and will not use it.
I believe this is important. The public who come to a talk is, most of
the time, here to learn about a piece of technology. They might have
some prior knowledge, but most of them don't. They want to learn what
can be done with it; how to use it; how it's an improvement. But more
importantly, we need to talk about the cost and path to adopt the piece
of technology. Going from a simple demo running on 2 hosts to a
something running on production with hundred of thousands of users and
on thousands of instances is a different story.
And yes, these could be questions asked by the public at the end of the
talk. But not everybody feel comfortable asking them out loud in front
of their peers.
I feel the same way about post-mortems. Companies should share them more
frequently. Some companies are
[[https://github.com/danluu/post-mortems][pretty good about it]]. I can
understand, if your product is not a service for developers, that you
might not want to share them on your blog to not scare your users. But
we should still share them during conferences. Maybe there's even an
opportunity for a meet-up focused on post-Mort em ?
Talking about issues and how difficult it might be to adopt something is
not doing is disservice to something you really enjoy working with.
|
