TRANSKRIP 🔗
WEBVTT
00:00:00.000 --> 00:00:03.020
Deployment was supposed to get easier. We got
00:00:03.020 --> 00:00:07.519
cloud platforms, containers, Kubernetes, Terraform,
00:00:07.780 --> 00:00:11.699
GitOps, internal developer platforms, and now
00:00:11.699 --> 00:00:15.380
AI agents that can generate code faster than
00:00:15.380 --> 00:00:18.500
most teams can review it. And yet, for a lot
00:00:18.500 --> 00:00:21.780
of teams, shipping software still feels weirdly
00:00:21.780 --> 00:00:25.399
painful. Not because people do not know how to
00:00:25.399 --> 00:00:28.929
deploy. Because deployment is where all the hidden
00:00:28.929 --> 00:00:32.289
complexity shows up. The app stack, the runtime,
00:00:32.649 --> 00:00:36.590
the registry, the secrets, the networking, the
00:00:36.590 --> 00:00:39.990
database, the cloud provider, the rollback path,
00:00:40.210 --> 00:00:43.750
the weird internal Jenkins job nobody wants to
00:00:43.750 --> 00:00:47.770
touch. And for smaller teams, that can be brutal.
00:00:47.929 --> 00:00:51.250
They do not always need every knob Kubernetes
00:00:51.250 --> 00:00:54.869
gives them. they do not always need a giant platform
00:00:54.869 --> 00:00:58.189
engineering program. Sometimes they just need
00:00:58.189 --> 00:01:02.649
a clean way to build, deploy, get logs, roll
00:01:02.649 --> 00:01:06.670
back, and move on. So maybe the future of deployment
00:01:06.670 --> 00:01:10.530
is not more flexibility. Maybe it is fewer decisions,
00:01:10.810 --> 00:01:14.109
better defaults, and picking the right opinions
00:01:14.109 --> 00:01:17.569
to outsource. I'm Brian Teller from Teller's
00:01:17.569 --> 00:01:38.019
Tech, and this is Ship It Weekly. Welcome back
00:01:38.019 --> 00:01:40.920
to Ship It Weekly, where I filter the noise and
00:01:40.920 --> 00:01:43.900
focus on what actually matters when you are the
00:01:43.900 --> 00:01:46.140
one running infrastructure and owning reliability.
00:01:46.700 --> 00:01:50.120
Most weeks, it's a quick news recap. In between
00:01:50.120 --> 00:01:53.200
those, I do conversation episodes with people
00:01:53.200 --> 00:01:55.819
who are building platforms, running infrastructure,
00:01:56.159 --> 00:01:59.159
organizing events, and thinking through where
00:01:59.159 --> 00:02:02.819
this industry is actually headed. Today is one
00:02:02.819 --> 00:02:05.799
of those conversations. I'm joined by Evan Phoenix,
00:02:06.140 --> 00:02:10.379
CEO of Miren. Evan previously worked on Terraform
00:02:10.379 --> 00:02:13.620
Enterprise and Waypoint at HashiCorp. And he
00:02:13.620 --> 00:02:16.780
also built Puma and Rubinius. We talk about why
00:02:16.780 --> 00:02:19.620
deployment is still painful. What teams keep
00:02:19.620 --> 00:02:22.740
getting wrong when they try to simplify it. And
00:02:22.740 --> 00:02:26.280
why small teams may not need more knobs. They
00:02:26.280 --> 00:02:29.139
may need better opinions. There is also a good
00:02:29.139 --> 00:02:33.110
thread in here around Terraform. OpenTofu, Terragrunt,
00:02:33.250 --> 00:02:37.430
Waypoint, platform abstractions, and what happens
00:02:37.430 --> 00:02:40.889
when tools make users do too much homework before
00:02:40.889 --> 00:02:44.610
they get value. We also get into AI because yes,
00:02:44.789 --> 00:02:47.849
AI can generate infrastructure code, but when
00:02:47.849 --> 00:02:50.990
that generated deployment setup breaks, who is
00:02:50.990 --> 00:02:54.550
on call for it? And does anyone actually understand
00:02:54.550 --> 00:02:58.069
why it was wired together that way? So if you
00:02:58.069 --> 00:03:01.349
work in DevOps, platform engineering, infrastructure,
00:03:01.650 --> 00:03:05.030
developer experience, or you have ever stared
00:03:05.030 --> 00:03:08.650
at a deployment system and thought, why is this
00:03:08.650 --> 00:03:11.889
still so annoying? This one should be worth your
00:03:11.889 --> 00:03:21.389
time. All right, let's jump in. Today, I'm joined
00:03:21.389 --> 00:03:24.509
by Evan Phoenix. He is the CEO of Miren, previously
00:03:24.509 --> 00:03:27.330
worked at Terraform Enterprise and Waypoint at
00:03:27.330 --> 00:03:30.479
HashiCorp. and he built Puma and Rubinius. We're
00:03:30.479 --> 00:03:33.460
talking about why deployment is so painful and
00:03:33.460 --> 00:03:36.539
what teams keep getting wrong when they try to
00:03:36.539 --> 00:03:38.560
simplify it. Evan, thank you for joining me.
00:03:38.840 --> 00:03:42.060
Oh, my pleasure. Glad to be here. So walk me
00:03:42.060 --> 00:03:45.180
through your history in tech. It's kind of interesting.
00:03:45.300 --> 00:03:48.379
I'm curious. How did you land where you are now?
00:03:48.599 --> 00:03:51.780
Well, I'm old enough that I graduated from college
00:03:51.780 --> 00:03:55.770
right after the bubble burst. The dot-com bubble
00:03:55.770 --> 00:03:58.409
burst. And so a lot of my career is actually
00:03:58.409 --> 00:04:02.349
informed by lack of opportunity post bubble bursting.
00:04:02.469 --> 00:04:05.150
I did all kinds of crazy interviews trying to
00:04:05.150 --> 00:04:08.870
just get any job. I almost wrote web Perl
00:04:08.870 --> 00:04:12.330
for many interviews and stuff like
00:04:12.330 --> 00:04:14.150
that. I couldn't get anything. And so I end up
00:04:14.150 --> 00:04:17.449
working for like a tiny ISP in Seattle. And then
00:04:17.449 --> 00:04:22.069
from there, I kind of just. was sort of scrapping
00:04:22.069 --> 00:04:24.389
i was like you know i met a guy in a hallway
00:04:24.389 --> 00:04:27.310
and he told me about this company he's working
00:04:27.310 --> 00:04:30.370
at and so i went to lunch and the lunch turned
00:04:30.370 --> 00:04:33.230
into a job interview and i worked at that company
00:04:33.230 --> 00:04:35.750
for a number of years and so like i didn't have
00:04:35.750 --> 00:04:37.810
any a moment where it was like oh there's this
00:04:37.810 --> 00:04:39.410
big thing and so
Load full transcript
i kind of was jumping around
00:04:39.410 --> 00:04:42.329
and the other out the other part of it was i
00:04:42.329 --> 00:04:46.259
started doing open source work in the late 90s
00:04:46.259 --> 00:04:49.220
at this point. And I did it all the way through
00:04:49.220 --> 00:04:52.019
college and then into my first few jobs. My first
00:04:52.019 --> 00:04:54.279
few jobs were, I was working at a three -person
00:04:54.279 --> 00:04:57.639
ISP. So I wasn't programming. I was like, open
00:04:57.639 --> 00:04:59.959
source is what I had for programming. And so
00:04:59.959 --> 00:05:03.980
I did that for a long time. And then I actually,
00:05:04.160 --> 00:05:06.279
the company I was working for in Seattle at the
00:05:06.279 --> 00:05:09.879
time was spinning down. I had a Ruby project
00:05:09.879 --> 00:05:13.040
and I actually got. hired by Engine Yard, for
00:05:13.040 --> 00:05:14.879
people who remember Engine Yard from the Ruby
00:05:14.879 --> 00:05:18.920
days, and to basically work on Rubinius, my Ruby
00:05:18.920 --> 00:05:23.079
implementation, and then also help out with sysadmin
00:05:23.079 --> 00:05:26.360
work. And I kept ignoring the sysadmin work and
00:05:26.360 --> 00:05:28.079
just working on Rubinius. And so they were like,
00:05:28.120 --> 00:05:29.959
I'll just do that. So I got to work on my open
00:05:29.959 --> 00:05:32.019
source project for like five years. They just
00:05:32.019 --> 00:05:33.920
basically paid me. It's like living the dream.
00:05:34.180 --> 00:05:38.310
And so like that. sort of balance like working
00:05:38.310 --> 00:05:40.370
on open source and then finding a job and finding
00:05:40.370 --> 00:05:41.930
interesting things and working on open source
00:05:41.930 --> 00:05:44.230
has been this sort of constant thing and then
00:05:44.230 --> 00:05:46.610
it it carried me through many jobs it carried
00:05:46.610 --> 00:05:50.029
me into uh a startup that i had before HashiCorp
00:05:50.029 --> 00:05:52.029
and then HashiCorp is a great sort of like job
00:05:52.029 --> 00:05:54.129
and open source thing that carried me all the
00:05:54.129 --> 00:05:56.370
way through that and then my current venture
00:05:56.370 --> 00:05:58.810
Miren is a similar sort of thing it's like job
00:05:58.810 --> 00:06:00.930
and open source and all kinds of things and so
00:06:00.930 --> 00:06:04.029
yeah it's been that's the that's the short long
00:06:04.029 --> 00:06:09.480
version Very cool. So my first job was, to date
00:06:09.480 --> 00:06:12.319
myself to a degree, was doing dial -up tech support.
00:06:12.560 --> 00:06:17.019
Yeah. Me too. That was the ISP. Yeah. I did dial
00:06:17.019 --> 00:06:20.329
-up tech support. I did DSL tech support. And
00:06:20.329 --> 00:06:23.870
then we were putting wireless antennas on people's
00:06:23.870 --> 00:06:26.230
houses around Seattle. For long range. Yeah.
00:06:26.389 --> 00:06:29.389
It was sort of long range. But yeah, directional.
00:06:29.389 --> 00:06:31.250
Yes, yes. But big antennas. That was actually
00:06:31.250 --> 00:06:33.449
how I got the, how that guy in the hallway talked
00:06:33.449 --> 00:06:35.829
to me was because I was carrying a giant antenna.
00:06:36.110 --> 00:06:38.750
And he was like, what is going on here? And I
00:06:38.750 --> 00:06:40.209
explained it to him. He's like, we should have
00:06:40.209 --> 00:06:41.910
lunch. And then, yeah, one thing or another.
00:06:42.110 --> 00:06:44.430
But yeah. That's awesome. So where does Miren
00:06:44.430 --> 00:06:46.850
fit in? What is Miren and how does it fit into
00:06:46.850 --> 00:06:49.329
the ecosystem? So, I mean, it's a software deployment
00:06:49.329 --> 00:06:53.029
system. It comes out of me constantly feeling
00:06:53.029 --> 00:06:55.569
like we are always struggling with deployment
00:06:55.569 --> 00:06:59.019
systems, right? Like people, I have... People
00:06:59.019 --> 00:07:01.959
have lots of great ideas and especially now with
00:07:01.959 --> 00:07:04.439
coding agents, you know, getting an idea out
00:07:04.439 --> 00:07:06.160
there and getting it like, oh, my gosh, I got
00:07:06.160 --> 00:07:07.680
this idea and I want to put the code together.
00:07:07.800 --> 00:07:09.420
And then you have to figure out, like, how do
00:07:09.420 --> 00:07:11.120
I give this to somebody? How do I let somebody
00:07:11.120 --> 00:07:13.680
else use this? It's always like it breaks you
00:07:13.680 --> 00:07:15.500
completely out of your rhythm. And so the like
00:07:15.500 --> 00:07:17.860
one of the original ideas was like, how do we
00:07:17.860 --> 00:07:20.259
integrate deployment? actually into sort of the
00:07:20.259 --> 00:07:22.620
software development life cycle right so that
00:07:22.620 --> 00:07:24.540
it's like it's not this thing where like i spent
00:07:24.540 --> 00:07:26.579
all this time and now like oh god now i got to
00:07:26.579 --> 00:07:28.259
deal with deployment but it's like oh deployment
00:07:28.259 --> 00:07:30.500
is part of that i go through a deploy i can get
00:07:30.500 --> 00:07:32.480
the logs it feels like it's part of the process
00:07:32.480 --> 00:07:35.740
and so i i've worked with Kubernetes i worked
00:07:35.740 --> 00:07:38.680
with Nomad i worked with lot most of the AWS
00:07:38.680 --> 00:07:41.639
products to do deployment and i felt like They're
00:07:41.639 --> 00:07:44.300
always missing something. And I wanted to feel
00:07:44.300 --> 00:07:47.819
like I had my own things to say about where that
00:07:47.819 --> 00:07:53.220
fit in. And the first cut of it is a self -hosted
00:07:53.220 --> 00:07:54.939
piece of software. So it's open source. You go
00:07:54.939 --> 00:07:56.740
and you run it wherever you want to run it. And
00:07:56.740 --> 00:07:58.620
so that makes it very accessible. That opens
00:07:58.620 --> 00:08:00.680
the doorway to lots of different people who have
00:08:00.680 --> 00:08:02.920
different needs. You want to data sovereignty.
00:08:03.240 --> 00:08:06.120
You want, oh, I'm really cost conscious. I want
00:08:06.120 --> 00:08:09.519
to run it in one particular way. And one of the
00:08:09.519 --> 00:08:13.040
big things. to do it for us was like, hey, you've
00:08:13.040 --> 00:08:16.079
got this open source. You can run the deployment
00:08:16.079 --> 00:08:17.939
system on your laptop. You can just deploy to
00:08:17.939 --> 00:08:19.699
your laptop. And then when you're going and you're
00:08:19.699 --> 00:08:21.720
deploying to production, it's the same. It's
00:08:21.720 --> 00:08:23.579
exactly the same. And so you've got all of these,
00:08:23.699 --> 00:08:26.360
all of your rhythm about how do I do it and how
00:08:26.360 --> 00:08:28.600
do I work is exactly the same. And so now you
00:08:28.600 --> 00:08:30.199
become that much more comfortable with the system
00:08:30.199 --> 00:08:33.200
and then you can sort of grow with it. That's
00:08:33.200 --> 00:08:35.940
awesome. That's great to hear. Okay, so. Why
00:08:35.940 --> 00:08:39.080
is deployment still such a mess after years of
00:08:39.080 --> 00:08:42.740
platforms and abstractions? Why? What's going
00:08:42.740 --> 00:08:45.940
on? Yeah, I mean, it really is the final boss
00:08:45.940 --> 00:08:49.879
in some ways of software. I think that it has
00:08:49.879 --> 00:08:51.779
to do with a few different things. I think that
00:08:51.779 --> 00:08:54.820
if we all were working in sort of one stack.
00:08:55.520 --> 00:08:58.500
one system, I think it would be a lot simpler.
00:08:58.659 --> 00:09:00.720
I don't think we'd be as happy, but I think that
00:09:00.720 --> 00:09:03.340
deployment would be simpler. And because then
00:09:03.340 --> 00:09:05.600
there would be just fewer knobs. There'd be,
00:09:05.620 --> 00:09:08.720
you know, like all the, you know, if Java, maybe,
00:09:08.779 --> 00:09:11.340
maybe not. I mean, it's not like Java beans really
00:09:11.340 --> 00:09:13.700
solved all of the problems of the, you know,
00:09:13.700 --> 00:09:17.460
like the late 90s, early 2000s. But there are
00:09:17.460 --> 00:09:19.519
so many different shapes that an application
00:09:19.519 --> 00:09:22.399
comes in. And some of that are just within the
00:09:22.399 --> 00:09:26.059
stack. One JavaScript app could be completely
00:09:26.059 --> 00:09:28.039
different and have a completely different profile
00:09:28.039 --> 00:09:31.259
than the next one. But then you multiply that
00:09:31.259 --> 00:09:33.620
matrix by all the different stacks and all the
00:09:33.620 --> 00:09:34.960
different things. And now you've got sort of
00:09:34.960 --> 00:09:38.840
a giant 3D cube of different ways that an app
00:09:38.840 --> 00:09:41.600
can behave. And I think that deployment systems
00:09:41.600 --> 00:09:46.210
try to say, OK, I want to do I want to give the
00:09:46.210 --> 00:09:48.889
power of all of these things all at once? And
00:09:48.889 --> 00:09:52.590
if so, I have to let you configure every possible
00:09:52.590 --> 00:09:55.049
thing on here so that I can have every possible
00:09:55.049 --> 00:09:58.250
stack here, right? And sometimes they're like,
00:09:58.330 --> 00:10:00.629
well, that's really too difficult. Most people
00:10:00.629 --> 00:10:02.649
are in this one quadrant. Let's actually focus
00:10:02.649 --> 00:10:05.370
on the kind of apps that look in this shape,
00:10:05.529 --> 00:10:08.649
your HTTP apps, request times under a second,
00:10:08.750 --> 00:10:10.730
whatever it might be. And great, let's sort of
00:10:10.730 --> 00:10:15.620
focus on that and try to distill down the needs
00:10:15.620 --> 00:10:17.759
and the function and that kind of thing and make
00:10:17.759 --> 00:10:21.440
it better. I think that, I do actually think
00:10:21.440 --> 00:10:24.419
it's better now than it used to be. I actually
00:10:24.419 --> 00:10:27.860
think there was a time where it was. better than
00:10:27.860 --> 00:10:31.019
it is now also but like it's just because of
00:10:31.019 --> 00:10:33.519
the number of variables that come into it and
00:10:33.519 --> 00:10:35.899
then what each platform each platform wants to
00:10:35.899 --> 00:10:37.779
have a unique say on it right so they want to
00:10:37.779 --> 00:10:40.039
say like oh well we've got this cool load balancer
00:10:40.039 --> 00:10:41.940
and so you should lean on the load balancer and
00:10:41.940 --> 00:10:43.519
now i'm like okay i gotta learn load balancer
00:10:43.519 --> 00:10:44.980
and i gotta learn all this kind of stuff right
00:10:44.980 --> 00:10:48.419
and so or uh you know AWS is like great for like
00:10:48.419 --> 00:10:51.419
oh we've got these AWS specific services to like
00:10:51.419 --> 00:10:54.080
make your app work in this one way of these different
00:10:54.080 --> 00:10:56.940
shapes and so it's the the amount of configuration
00:10:56.940 --> 00:11:00.820
the amount of of of functionality that you want
00:11:00.820 --> 00:11:03.259
to sort of eek from the platform too you think
00:11:03.259 --> 00:11:07.700
ai helps us bridge that gap better i mean in
00:11:07.700 --> 00:11:10.740
some ways yes in some ways because it's really
00:11:10.740 --> 00:11:13.980
good ai is really good at if you could if a platform
00:11:13.980 --> 00:11:16.470
can explain itself The AI is really good at being
00:11:16.470 --> 00:11:18.450
like, okay, great. The platform works like this.
00:11:18.769 --> 00:11:21.629
I'm going to basically match the app to the platform,
00:11:21.789 --> 00:11:23.389
right? So the platform can have a bunch of different
00:11:23.389 --> 00:11:25.429
things it does well, things that it doesn't do
00:11:25.429 --> 00:11:27.470
well. And the AI can say like, okay, I'm going
00:11:27.470 --> 00:11:29.690
to make the app work. You wanted this platform.
00:11:29.889 --> 00:11:32.009
Great. I'm going to plug you into it. The inverse
00:11:32.009 --> 00:11:35.429
is also true, although I don't think we're seeing
00:11:35.429
--> 00:11:39.590
it that much yet. The app exists as this thing,
00:11:39.610 --> 00:11:41.669
and the AI is like, great, let me find the deployment
00:11:41.669 --> 00:11:43.490
platform that fits this. I don't think we're
00:11:43.490 --> 00:11:46.690
seeing that quite yet, maybe eventually. I do
00:11:46.690 --> 00:11:52.409
think it's making it better. Okay, so you've
00:11:52.409 --> 00:11:54.669
lived through multiple waves of this will fix
00:11:54.669 --> 00:11:58.309
it. What did people misunderstand about Terraform
00:11:58.309 --> 00:12:01.330
Enterprise early on? You know, we were constantly
00:12:01.330 --> 00:12:06.399
trying to figure out the shape of it. There's
00:12:06.399 --> 00:12:09.980
a whole Conway's Law situation with Terraform
00:12:09.980 --> 00:12:12.059
Enterprise, right? So we were constantly trying
00:12:12.059 --> 00:12:13.679
to help people figure out the shape of like,
00:12:13.759 --> 00:12:17.080
what should their repos look like? So in other
00:12:17.080 --> 00:12:19.779
words, like, should it be like there's a repo
00:12:19.779 --> 00:12:23.259
called Infra and it has like a thousand modules
00:12:23.259 --> 00:12:25.899
that are basically going and configuring all
00:12:25.899 --> 00:12:28.740
these different things. And every time anybody
00:12:28.740 --> 00:12:31.139
changes anything, we have to run a plan that
00:12:31.139 --> 00:12:34.470
basically touches like, you know. 1 ,500 resources
00:12:34.470 --> 00:12:37.269
or something like that, right? Do you want that?
00:12:37.370 --> 00:12:41.269
Or do you want like, oh, well, I want my, like,
00:12:41.330 --> 00:12:43.590
should it be tiered? So it's like, okay, the
00:12:43.590 --> 00:12:46.450
load balancers are a module, a thing, there's
00:12:46.450 --> 00:12:49.990
like infra -loadlb, and that's just the load
00:12:49.990 --> 00:12:54.049
balancers. And so it was, and I don't know that
00:12:54.049 --> 00:12:57.500
Terraform really ever figured this out. even
00:12:57.500 --> 00:13:01.200
to this day, if I'm honest. But people were constantly,
00:13:01.360 --> 00:13:04.419
that was the big question. We had white glove
00:13:04.419 --> 00:13:06.399
service at HashiCorp at one point that was
00:13:06.399 --> 00:13:09.879
like, how do I organize my Terraform modules,
00:13:10.139 --> 00:13:13.340
right? In a way that's like, that I can apply
00:13:13.340 --> 00:13:15.460
them quickly, that reduces the blast radius,
00:13:15.539 --> 00:13:17.960
all of those sorts of things. And so I think
00:13:17.960 --> 00:13:21.220
that was a really interesting thing that, and
00:13:21.220 --> 00:13:22.679
that wasn't Terraform Enterprise specifically,
00:13:22.700 --> 00:13:24.779
that was Terraform writ large. I think Terraform
00:13:24.779 --> 00:13:27.440
Enterprise was a, The first version of Terraform
00:13:27.440 --> 00:13:29.639
Enterprise was the one that I worked on where
00:13:29.639 --> 00:13:36.440
it was for people who wanted the experience of
00:13:36.440 --> 00:13:39.360
having a platform that ran Terraform for them,
00:13:39.399 --> 00:13:42.019
but they needed to have it sort of inside their
00:13:42.019 --> 00:13:45.100
silo, whatever that meant. And so we went through
00:13:45.100 --> 00:13:47.399
a few iterations where basically it was just
00:13:47.399 --> 00:13:50.919
like a – big machine like we scaled it vertically
00:13:50.919 --> 00:13:53.539
inside the actual instance and then you would
00:13:53.539 --> 00:13:55.980
just it would just run Terraform on on instance
00:13:55.980 --> 00:14:01.519
and it was a like so many products when you see
00:14:01.519 --> 00:14:03.580
how the sausage is made underneath you're like
00:14:03.580 --> 00:14:06.059
this is all we're doing you know like it was
00:14:06.059 --> 00:14:08.299
a little bit that in a good way because it basically
00:14:08.299 --> 00:14:10.879
made it made the product easy to manage and and
00:14:10.879 --> 00:14:12.419
simple and stuff like that like we were constantly
00:14:12.419 --> 00:14:14.779
like stripping things out and taking things out
00:14:14.779 --> 00:14:17.379
and like you know at one point the product was
00:14:17.379 --> 00:14:20.200
dependent upon RabbitMQ and we're like, no, we
00:14:20.200 --> 00:14:23.340
need to like tear, like slice this down. Like
00:14:23.340 --> 00:14:25.860
that's too big of a complexity for a thing that's
00:14:25.860 --> 00:14:29.000
just running on a single machine. And so. Is
00:14:29.000 --> 00:14:31.039
that for like worker queues? Yeah, for worker
00:14:31.039 --> 00:14:33.480
queues. Because that was, so at the time, this
00:14:33.480 --> 00:14:37.720
is Terraform Enterprise is a product before
00:14:37.720 --> 00:14:41.039
Terraform Cloud, which is a weird sequencing
00:14:41.039 --> 00:14:44.379
because we had Atlas at the time, which was the
00:14:44.799 --> 00:14:47.519
thing that predated HCP which was like our
00:14:47.519 --> 00:14:50.879
all-in-one platform and uh Terraform Enterprise
00:14:50.879 --> 00:14:53.120
existed as sort of a side thing to atlas and
00:14:53.120 --> 00:14:55.399
then atlas eventually gets broken apart part
00:14:55.399 --> 00:14:57.279
of it became Terraform Cloud we still have
00:14:57.279 --> 00:15:00.799
uh Terraform Enterprise and so that rabbit mq
00:15:00.799 --> 00:15:04.740
part came from the original vision of atlas yeah
00:15:04.740 --> 00:15:07.059
believe it or not i was an early uh customer
00:15:07.059 --> 00:15:11.000
to Terraform Enterprise yeah i we so probably
00:15:11.000 --> 00:15:13.399
much like a lot of the customers we we had a
00:15:13.399 --> 00:15:16.919
pci HIPAA requirement where we needed to have
00:15:16.919 --> 00:15:19.000
state files managed locally. We needed to make
00:15:19.000 --> 00:15:20.679
sure that everything was self -contained and
00:15:20.679 --> 00:15:23.539
then eventually move to Terraform Cloud. Did
00:15:23.539 --> 00:15:28.399
you use it in the... You weren't at desk.com,
00:15:28.500 --> 00:15:31.740
were you? It was not, no. Was it a forms company
00:15:31.740 --> 00:15:35.480
at the time? There's three shapes of Terraform
00:15:35.480 --> 00:15:38.120
Enterprise. The first, the shape number one was
00:15:38.120 --> 00:15:40.759
only ever seen by like two customers. Oh, okay.
00:15:40.799 --> 00:15:42.720
One of which was desk.com when they got bought
00:15:42.720 --> 00:15:46.139
by Salesforce. And then the second one was the
00:15:46.139 --> 00:15:49.120
AWS only Terraform Enterprise where we basically
00:15:49.120 --> 00:15:51.379
shared an AMI and a bunch of Terraform for you
00:15:51.379 --> 00:15:55.909
to run it. That was shape number two. Okay. And
00:15:55.909 --> 00:15:58.649
then shape number three was a rewrite that I
00:15:58.649 --> 00:16:01.750
did, which made it basically Docker -based. And
00:16:01.750 --> 00:16:04.009
then you can install it anywhere. I forget the
00:16:04.009 --> 00:16:06.110
name of the application that it was built on.
00:16:06.809 --> 00:16:10.149
Replicated. Replicated, yes. I remember going
00:16:10.149 --> 00:16:13.370
through and dealing with weird Docker IP address
00:16:13.370 --> 00:16:17.529
allocation issues. All the time, yes. So yeah,
00:16:17.570 --> 00:16:19.049
I guess it was version three. But yeah, it was
00:16:19.049 --> 00:16:22.230
a good product. We eventually moved to Terraform
00:16:22.230 --> 00:16:24.960
Cloud. And then I think they've since, I don't
00:16:24.960 --> 00:16:26.580
know what the company does now. It was a few
00:16:26.580 --> 00:16:30.379
years ago. But it got me excited about IaC because
00:16:30.379 --> 00:16:33.899
at the time, and this is going to cause a lot
00:16:33.899 --> 00:16:36.299
of people to cringe, I was at a company where
00:16:36.299 --> 00:16:39.399
we were using Ansible with no state. We were
00:16:39.399 --> 00:16:44.220
just using Boto3 to manage infrastructure, some
00:16:44.220 --> 00:16:48.750
CDK stuff. So going from that to Terraform was
00:16:48.750 --> 00:16:52.169
a huge jump. More recently, the problem you were
00:16:52.169 --> 00:16:55.730
describing earlier, just about how do you manage
00:16:55.730 --> 00:16:57.309
state files, how do you manage environments,
00:16:57.509 --> 00:17:01.049
how do you manage modules, I feel like Terragrunt
00:17:01.049 --> 00:17:05.250
has really tried to answer that problem. I don't
00:17:05.250 --> 00:17:07.650
know that they've answered it well, to be honest
00:17:07.650 --> 00:17:09.490
with you, but I think they've tried at least
00:17:09.490 --> 00:17:12.779
to answer some of that dry issue. Yeah. I mean,
00:17:12.779 --> 00:17:14.920
they were around right when I was at HashiCorp
00:17:14.920 --> 00:17:18.500
because the Terragrunt has been around for maybe
00:17:18.500 --> 00:17:20.579
not 10 years, but quite a while at this point.
00:17:20.700 --> 00:17:22.940
Yeah. And they had this idea of like, how do
00:17:22.940 --> 00:17:24.980
you slice your Terraform and
we'll basically
00:17:24.980 --> 00:17:28.099
like put it in different shapes. And they were
00:17:28.099 --> 00:17:30.819
very successful at basically saying, we're not
00:17:30.819 --> 00:17:32.920
going to let you like Terraform in some ways,
00:17:32.960 --> 00:17:35.380
let you do anything, let you put files wherever
00:17:35.380 --> 00:17:37.000
and let you organize it. However, in Terragrunt
00:17:37.000 --> 00:17:39.019
was like, what if I just put like a corral, like
00:17:39.019 --> 00:17:41.769
I put like a little, like a little. guard around
00:17:41.769 --> 00:17:44.349
this thing you can't do all this stuff it's just
00:17:44.349 --> 00:17:46.690
like it's going to work like this and and that
00:17:46.690 --> 00:17:49.329
is very powerful from a workflows perspective
00:17:49.329 --> 00:17:52.809
because now people are they're having to comport
00:17:52.809 --> 00:17:55.769
themselves maybe to this smaller feature set
00:17:55.769 --> 00:17:57.470
but they're also not having to think as much
00:17:57.470 --> 00:17:59.029
because they're basically being like no it works
00:17:59.029 --> 00:18:00.609
like this like i'm supposed to supposed to put
00:18:00.609 --> 00:18:02.750
it in these boxes we think about that a lot about
00:18:02.750 --> 00:18:06.000
that at Miren because Part of it is what do you
00:18:06.000 --> 00:18:08.220
want someone to have an opinion about? Like Terraform
00:18:08.220 --> 00:18:11.559
made you have opinions about like file structure
00:18:11.559 --> 00:18:14.519
and stuff like that. Right. Terragrunt is all
00:18:14.519 --> 00:18:16.539
about like, OK, well, let's not like people have
00:18:16.539 --> 00:18:18.319
an opinion about all of these things. Let's just
00:18:18.319 --> 00:18:20.200
make those opinions for them and we'll just tell
00:18:20.200 --> 00:18:22.400
them what they are. And then they're not having
00:18:22.400 --> 00:18:24.619
to go around, go about, figure out, like, should
00:18:24.619 --> 00:18:26.640
I have an opinion about this? Like a lot of people,
00:18:26.720 --> 00:18:28.619
when they would start using Terraform, they're
00:18:28.619 --> 00:18:30.619
like, but just can you just tell me what the
00:18:30.619 --> 00:18:32.420
right pattern is? And they're like, well, like,
00:18:32.500 --> 00:18:34.819
did you do that? or does it think this or what you
00:18:34.819 --> 00:18:38.099
know whatever and so uh they would we would sort
00:18:38.099 --> 00:18:40.200
of force them in some ways to have an opinion
00:18:40.200 --> 00:18:41.779
about stuff that they didn't really want to have
00:18:41.779 --> 00:18:45.460
an opinion about that's fair so where do you
00:18:45.460 --> 00:18:50.319
think OpenTofu sits in that opinionated stack
00:18:50.319 --> 00:18:54.319
against Terraform Terragrunt yeah at least whatever
00:18:54.319 --> 00:18:56.720
other alternatives there are well i think open
00:18:56.720 --> 00:19:01.630
tofu is the result of two specific things obviously
00:19:01.630
--> 00:19:04.470
number one was a license change without a question
00:19:04.470 --> 00:19:07.650
no no no question right but i think that the
00:19:07.650 --> 00:19:11.390
second one is part of it which is that the backlog
00:19:11.390 --> 00:19:15.430
of potential features for Terraform that we're
00:19:15.430 --> 00:19:18.289
sitting in PRs or sitting in issues that the
00:19:18.289 --> 00:19:19.769
Terraform team was like no we're not going to
00:19:19.769 --> 00:19:21.650
do that no we're not going to do that and so
00:19:21.650 --> 00:19:27.069
the the uh The license change catalyzed people
00:19:27.069 --> 00:19:28.990
to be like, okay, well, I want a different thing.
00:19:29.170 --> 00:19:32.369
And then it got runway because it was like, well,
00:19:32.430 --> 00:19:34.170
look at all these PRs. Look at all these really
00:19:34.170 --> 00:19:36.470
cool features that Terraform team never wanted
00:19:36.470 --> 00:19:38.630
to do that will just do those things, right?
00:19:38.710 --> 00:19:40.529
And so now it's like it's growing all these different,
00:19:40.650 --> 00:19:43.690
growing in this sort of different direction because
00:19:43.690 --> 00:19:47.849
of that, right? It might still exist even if
00:19:47.849 --> 00:19:49.890
there wasn't all those features, but I don't
00:19:49.890 --> 00:19:52.349
think it would have had the legs that it has
00:19:52.349 --> 00:19:55.069
had. if it didn't have that second part, right?
00:19:55.269 --> 00:19:58.230
And I think that what's interesting is that I
00:19:58.230 --> 00:20:01.470
was at HashiCorp during the creation of OpenTofu.
00:20:01.549 --> 00:20:03.109
So we were talking about it and we were trying
00:20:03.109 --> 00:20:05.029
to figure out what it meant. And what was interesting
00:20:05.029 --> 00:20:07.529
is that OpenTofu was taking on these features
00:20:07.529 --> 00:20:12.170
that Terraform was rejecting. for good reason
00:20:12.170 --> 00:20:14.190
right so there was a lot of interesting features
00:20:14.190 --> 00:20:16.529
there's encryption encrypted state management
00:20:16.529 --> 00:20:18.549
was one of the big things that OpenTofu took
00:20:18.549 --> 00:20:22.269
on that Terraform was like don't do this do not
00:20:22.269 --> 00:20:23.990
like the reason that they didn't do it for the
00:20:23.990 --> 00:20:26.549
longest time was because like we we run the math
00:20:26.549 --> 00:20:30.730
and if you do this what you end up doing is that
00:20:30.730 --> 00:20:34.119
you create like this other big hairy availability
00:20:34.119 --> 00:20:37.880
scenario that is now i want to do stuff and now
00:20:37.880 --> 00:20:40.220
i've lost the keys to do stuff and now i can't
00:20:40.220 --> 00:20:42.460
do anything i'm sort of like really dead in the
00:20:42.460 --> 00:20:44.480
water and so they were like an advocate for like
00:20:44.480 --> 00:20:47.140
do it a different way right they had all these
00:20:47.140 --> 00:20:50.059
different things and so and again like maybe
00:20:50.059 --> 00:20:52.119
that's an example of where Terraform was like
00:20:52.119 --> 00:20:53.740
actually don't let people do everything they
00:20:53.740 --> 00:20:55.809
were trying to sort of rein in the the sort of
00:20:55.809 --> 00:20:57.930
field of opinion and and in some ways OpenTofu
00:20:57.930 --> 00:21:00.809
is like even wider like now you can do all these
00:21:00.809 --> 00:21:02.670
things and maybe some of them are good ideas
00:21:02.670 --> 00:21:05.450
but at least people are trying them out so well
00:21:05.450 --> 00:21:07.930
in secrets probably shouldn't be in state right
00:21:07.930 --> 00:21:11.309
or anything too sensitive or if it is i mean
00:21:11.309 --> 00:21:14.670
lock it down otherwise but but maybe that's the
00:21:14.670 --> 00:21:16.890
answer like i'm sure they were pushing like Vault
00:21:16.890 --> 00:21:19.690
and like using yeah that was really that was
00:21:19.690 --> 00:21:21.490
that part of it was really interesting because
00:21:21.490 --> 00:21:24.670
that was the sort of the evolution of how of
00:21:24.670 --> 00:21:30.210
your state files as definitive reproducible data
00:21:30.619 --> 00:21:32.819
right so like one of the reasons early on in
00:21:32.819 --> 00:21:35.519
Terraform that Terraform would end up having
00:21:35.519 --> 00:21:37.640
secrets inside of it was because they were like
00:21:37.640 --> 00:21:39.920
look like the state file needs to be the thing
00:21:39.920 --> 00:21:43.920
the the the the jewel that you can use to recreate
00:21:43.920 --> 00:21:48.640
whatever you need right and so they they were
00:21:48.640 --> 00:21:50.339
like you should put secrets in your state file
00:21:50.339 --> 00:21:52.359
because you want to do this and the other thing
00:21:52.359 --> 00:21:55.460
and then what they sort of realized over time
00:21:55.460 --> 00:21:58.240
was that like that's a that's a hard position
00:21:58.240 --> 00:22:02.000
to to have defense for when Terraform depends
00:22:02.000 --> 00:22:05.440
on a jillion other APIs. And so it's like, okay,
00:22:05.500 --> 00:22:07.640
well, if it already depends on all these other
00:22:07.640 --> 00:22:10.319
APIs to work anyway, can't you just make secrets
00:22:10.319 --> 00:22:12.640
an API that you have to depend on? And then that
00:22:12.640 --> 00:22:14.380
became a thing that they sort of moved towards.
00:22:14.519 --> 00:22:16.769
In my opinion, they didn't move towards it. quick
00:22:16.769 --> 00:22:18.890
enough like in other words i i always felt like
00:22:18.890 --> 00:22:21.029
hey if you're gonna have secrets just make make
00:22:21.029 --> 00:22:23.089
a secrets part of Terraform Cloud and let secrets
00:22:23.089 --> 00:22:25.549
live there your state files can basically just
00:22:25.549 --> 00:22:27.069
like not have secrets in them no matter where
00:22:27.069 --> 00:22:29.170
you are you can just sort of outsource the secret
00:22:29.170 --> 00:22:31.750
management part to Terraform Cloud um that was
00:22:31.750 --> 00:22:33.609
kind of the idea with variable sets though right
00:22:33.609 --> 00:22:36.029
i mean yeah whether it's exactly right that's
00:22:36.029 --> 00:22:38.869
exactly right yes that's exactly right but it
00:22:38.869 --> 00:22:41.049
could have been at a deeper integration level
00:22:41.049 --> 00:22:44.240
no right so yeah for sure So what did you learn
00:22:44.240 --> 00:22:46.880
about what devs want versus what they think they
00:22:46.880 --> 00:22:50.440
want when it comes to tools like Waypoint? Yeah,
00:22:50.460 --> 00:22:54.220
well, Mitchell and I started Waypoint in 2019,
00:22:54.619 --> 00:22:57.759
which was an interesting time to start a new
00:22:57.759 --> 00:23:01.819
product because right in late 2019. And I think
00:23:01.819 --> 00:23:04.960
that the thing that we were really focused on
00:23:04.960 --> 00:23:09.069
was a control plane. for backend deployment systems.
00:23:09.390 --> 00:23:11.109
So it would be like you would talk to Waypoint,
00:23:11.289 --> 00:23:13.089
and Waypoint would go off and do things like,
00:23:13.230 --> 00:23:15.730
oh, to do that, I need to build a container.
00:23:15.869 --> 00:23:17.930
Let me go build a container. And then, OK, great,
00:23:18.009 --> 00:23:19.470
I will do this, and I will do this other thing,
00:23:19.549 --> 00:23:22.049
and I will do this third thing. And the idea
00:23:22.049 --> 00:23:25.529
was it was just a pure workflow engine for deployment.
00:23:25.730 --> 00:23:29.470
And it farmed out all of its backends to do different
00:23:29.470 --> 00:23:32.470
things. And one of the things that I learned
00:23:32.470 --> 00:23:36.160
working on that was You're asking a lot of someone
00:23:36.160 --> 00:23:39.440
who wants to use this thing to say it doesn't
00:23:39.440 --> 00:23:41.480
do anything until you do all of this homework
00:23:41.480 --> 00:23:44.099
first. Right. It was a lot of homework to get
00:23:44.099 --> 00:23:48.519
to get Waypoint running. And that kind of sucked.
00:23:48.619 --> 00:23:50.980
Right. Like from just from a usability perspective.
00:23:51.039 --> 00:23:54.420
But it also it wasn't great from like we're working
00:23:54.420 --> 00:23:57.460
on a perspective. One of the things that we we
00:23:57.460 --> 00:24:01.220
found out that was like a like a really big hurdle
00:24:01.220 --> 00:24:04.329
was people. people would get to the point where
00:24:04.329 --> 00:24:06.269
they're like, oh, like, oh, I just saw this on
00:24:06.269 --> 00:24:07.869
a system and it can run, it can build Docker
00:24:07.869 --> 00:24:10.049
containers. And that part was easy because we
00:24:10.049 --> 00:24:12.490
would just say, also run Docker on the same machine
00:24:12.490 --> 00:24:13.930
that you want to run Waypoint. Like that was
00:24:13.930 --> 00:24:15.910
sort of, we could easily pave over that. But
00:24:15.910 --> 00:24:17.829
then they'd be like, okay, now I want to deploy
00:24:17.829 --> 00:24:19.549
this to ECS. And we'd be like, okay, well, you
00:24:19.549 --> 00:24:21.829
got to go. You have to go configure, like, container
00:24:21.829 --> 00:24:23.829
registry, and then you've got to push it to the
00:24:23.829 --> 00:24:25.970
configure push to push the container registry,
00:24:26.089 --> 00:24:27.549
and then you've got to tell the container registry
00:24:27.549 --> 00:24:29.670
over here to pull. And it was like just that,
00:24:29.769 --> 00:24:31.930
just getting over the, like, where does this
00:24:31.930 --> 00:24:36.730
image go was a leap. It was a big hurdle for
00:24:36.730 --> 00:24:38.930
people to get over. Sometimes it was too big.
00:24:39.009 --> 00:24:40.630
Like, they wouldn't even get past that point.
00:24:40.869 --> 00:24:45.500
And so we worked on it to try to. to pave over
00:24:45.500 --> 00:24:47.779
it. One of the things that I did was I made it
00:24:47.779 --> 00:24:50.059
so that if you were using, let's say, ECS as
00:24:50.059 --> 00:24:52.039
an example, we would automatically configure
00:24:52.039 --> 00:24:58.099
a repo inside AWS's OCI registry to push those
00:24:58.099 --> 00:25:00.579
images to. So we would kind of just try to pave
00:25:00.579 --> 00:25:02.900
over that as much as we could. But the learning
00:25:02.900 --> 00:25:06.140
experience I took away from it was like, a pure
00:25:06.140 --> 00:25:11.390
workflow engine, it's not enough product. It's
00:25:11.390 --> 00:25:14.509
not enough unless there's this really complicated
00:25:14.509 --> 00:25:17.210
thing that people are doing that you're like,
00:25:17.250 --> 00:25:18.869
I want to come in and I want to just revolutionize
00:25:18.869 --> 00:25:21.349
this one segment with a workflow engine. Great.
00:25:21.670 --> 00:25:25.309
But deployment wasn't that. And what they really
00:25:25.309 --> 00:25:27.269
needed was they needed answers. It didn't need
00:25:27.269 --> 00:25:29.109
to be all the answers, but they needed a lot
00:25:29.109 --> 00:25:31.150
of answers that were already set in stone at
00:25:31.150 --> 00:25:32.529
the beginning. So when I was going to deploy
00:25:32.529 --> 00:25:34.109
stuff, I was like, it just deployed. It just
00:25:34.109 --> 00:25:37.190
did the thing. And part of that comes out of
00:25:37.190 --> 00:25:41.329
the idea of like. again, from HashiCorp that
00:25:41.329 --> 00:25:44.730
we learned was people wanted us to have opinions,
00:25:45.009 --> 00:25:47.250
right? They really wanted HashiCorp's opinions.
00:25:47.349 --> 00:25:48.569
They want to say like, well, okay, well, like
00:25:48.569 --> 00:25:50.609
great, it's Terraform, it's Vault, it's Nomad,
00:25:50.690 --> 00:25:52.109
but how do I set it up? Like, what's the best
00:25:52.109 --> 00:25:53.789
way? All this kind of stuff. They'd constantly
00:25:53.789 --> 00:25:55.970
be asking for our opinions, even though we had
00:25:55.970 --> 00:25:58.869
these sort of generic
tools. And so the takeaway
00:25:58.869 --> 00:26:00.569
that I took from it was like, well, let's just
00:26:00.569 --> 00:26:02.829
build a really opinionated tool, right? Let's
00:26:02.829 --> 00:26:05.460
go in and say, hey, we've got opinions. We've
00:26:05.460 --> 00:26:07.119
been working in the field for all this time.
00:26:07.259 --> 00:26:09.079
We've got opinions about how this stuff should
00:26:09.079 --> 00:26:10.519
work and what it should do and how it should
00:26:10.519 --> 00:26:13.740
feel and what the primitive should be. And we
00:26:13.740 --> 00:26:15.880
think that if you agree with those things, you
00:26:15.880 --> 00:26:17.240
will really like our thing because it just does
00:26:17.240 --> 00:26:19.259
all those things, right? It's sort of really
00:26:19.259 --> 00:26:22.420
defining that in -group of people who want something
00:26:22.420 --> 00:26:25.400
that looks and feels in a specific way. And so
00:26:25.400 --> 00:26:28.490
we... When we were working on Miren and from
00:26:28.490 --> 00:26:30.289
Waypoint, we took that opposite approach. We're
00:26:30.289 --> 00:26:32.190
like, let's just build all the pieces. Let's
00:26:32.190 --> 00:26:33.430
build the thing that builds stuff. That builds
00:26:33.430 --> 00:26:35.450
the thing that deploys it. Let's not tell people
00:26:35.450 --> 00:26:37.930
where stuff's being stored, where the images
00:26:37.930 --> 00:26:40.730
are. Let's just make that not a problem at all
00:26:40.730 --> 00:26:42.250
that people have to think about and worry about.
00:26:42.730 --> 00:26:45.549
And that was a big takeaway from Waypoint, right?
00:26:46.410 --> 00:26:51.730
Waypoint, it never really got traction and it
00:26:51.730 --> 00:26:53.829
really got its legs underneath it. Partly, I
00:26:53.829 --> 00:26:55.789
think, because it didn't have enough opinions.
00:26:56.319 --> 00:26:58.539
It didn't have enough opinions about how someone
00:26:58.539 --> 00:27:01.880
should actually use this thing. And then just
00:27:01.880 --> 00:27:03.460
because of things that were happening inside
00:27:03.460 --> 00:27:05.799
HashiCorp at the time, it ends up sort of
00:27:05.799 --> 00:27:08.980
migrating over to HCP and then it migrates away
00:27:08.980 --> 00:27:10.900
from deployment entirely. It migrates basically
00:27:10.900 --> 00:27:13.819
to a Terraform front end. And it wasn't doing
00:27:13.819 --> 00:27:15.460
deployment at all towards the end. It was sort
00:27:15.460 --> 00:27:20.900
of misnamed in my opinion. But yeah, that was
00:27:20.900 --> 00:27:24.819
the big takeaway from those Waypoint-as-deployment
00:27:24.819 --> 00:27:28.750
system days. Very cool. So, OK, jumping back
00:27:28.750 --> 00:27:32.410
to Miren. And I don't want to turn this into
00:27:32.410 --> 00:27:37.529
a product pitch, but I am curious what what I
00:27:37.529 --> 00:27:39.130
do want to understand is what you're trying to
00:27:39.130 --> 00:27:42.549
fix and like what's broken
for small teams trying
00:27:42.549 --> 00:27:44.849
to ship. What tradeoffs are you making to keep
00:27:44.849 --> 00:27:48.089
it simple? Yeah, I mean, I think that I'll start
00:27:48.089 --> 00:27:50.210
on the first one, right? Like what is the what
00:27:50.210 --> 00:27:53.319
makes Miren sort of interesting is that. And
00:27:53.319 --> 00:27:55.819
where do people and small teams hit problems,
00:27:56.039 --> 00:27:59.819
right? I think that as I've been looking at deployment
00:27:59.819 --> 00:28:03.740
tools for years, you sort of have stuff that's
00:28:03.740 --> 00:28:05.579
really good for a single developer. Like, oh,
00:28:05.619 --> 00:28:07.839
I've got Docker on a machine somewhere, and I
00:28:07.839 --> 00:28:11.880
just SSH stuff to it or whatever, right? And
00:28:11.880 --> 00:28:14.059
then there's sort of a gulf. And then you have
00:28:14.059 --> 00:28:16.920
like, OK, well, I want like a Kubernetes -based
00:28:16.920 --> 00:28:19.700
thing that's sort of eaten most of the market,
00:28:19.799 --> 00:28:23.690
right? There's not really much in that space
00:28:23.690 --> 00:28:26.210
for like, actually, I'm a team of five people
00:28:26.210 --> 00:28:28.809
that really just wants to like, that are application
00:28:28.809 --> 00:28:30.609
focused. They're just like, I just want to ship
00:28:30.609 --> 00:28:32.490
an application. I want to be able to deploy it
00:28:32.490 --> 00:28:35.809
30 times a day if I need to, right? And there's
00:28:35.809 --> 00:28:39.470
kind of not a lot there. There's a number of
00:28:39.470 --> 00:28:42.049
products that try to take what Kubernetes is
00:28:42.049 --> 00:28:46.710
and build facades and things inside Kubernetes
00:28:46.710 --> 00:28:48.759
in order to get. back to sort of this middle
00:28:48.759 --> 00:28:50.579
ground of like a thing that would feel good for
00:28:50.579 --> 00:28:53.660
a small team. But what small teams always run
00:28:53.660 --> 00:28:57.720
into is now you have not this much stuff, like
00:28:57.720 --> 00:28:59.440
from Kubernetes to where they were, but like
00:28:59.440 --> 00:29:02.440
this much stuff, my arm is way over there because
00:29:02.440 --> 00:29:05.200
now it's like, it's not the, if something goes
00:29:05.200 --> 00:29:07.559
wrong, it's not necessarily in this top layer.
00:29:07.720 --> 00:29:09.799
It's basically somewhere deep inside the Kubernetes
00:29:09.799 --> 00:29:12.980
stack of like your stateful pod had got evicted
00:29:12.980 --> 00:29:14.779
because the memory pressure was too high, whatever
00:29:14.779 --> 00:29:18.240
it is. Right. Yeah. And, we didn't want people
00:29:18.240 --> 00:29:19.880
to have to worry about that because that team
00:29:19.880 --> 00:29:21.480
of five people, they just want to deploy stuff.
00:29:21.640 --> 00:29:23.700
They want to get in the groove of deploying.
00:29:23.920 --> 00:29:25.420
They need something that really is getting out
00:29:25.420 --> 00:29:27.119
of their
way that just works the way they need
00:29:27.119 --> 00:29:31.089
it to work. The question was, how do you build
00:29:31.089 --> 00:29:33.150
something that isn't going to have a bunch of
00:29:33.150 --> 00:29:36.250
landmines, right? Like a lot of the deep Kubernetes
00:29:36.250 --> 00:29:38.269
stuff that's really important for big teams are
00:29:38.269 --> 00:29:41.869
landmines for small teams. And the answer was
00:29:41.869 --> 00:29:43.630
like, let's just strip it away. Let's just keep
00:29:43.630 --> 00:29:45.970
stripping stuff away until we get to something
00:29:45.970 --> 00:29:48.549
that feels like it's the right size for a small
00:29:48.549 --> 00:29:51.490
team. And that meant like, yeah, it's not Kubernetes
00:29:51.490 --> 00:29:53.809
-based at all. We were like, we're stripping
00:29:53.809 --> 00:29:56.170
that away entirely. We're like, okay, what's
00:29:56.170 --> 00:29:58.420
the right? what's the right primitive like containers
00:29:58.420 --> 00:29:59.960
are still the right primitive again you feel
00:29:59.960 --> 00:30:01.819
see where Docker is you see where people are
00:30:01.819 --> 00:30:04.440
writ large containers are still that thing but
00:30:04.440 --> 00:30:07.099
then it was like okay great let's say containers
00:30:07.099 --> 00:30:08.539
as a primitive and let's just start building
00:30:08.539 --> 00:30:10.140
on top of them and so there's things that miren
00:30:10.140 --> 00:30:12.700
does that Kubernetes does thousands of times
00:30:12.700 --> 00:30:16.059
better right that's no question right but i think
00:30:16.059 --> 00:30:18.519
that A lot of the things that Miren does, you
00:30:18.519 --> 00:30:20.700
could do with Kubernetes, but Miren does it out
00:30:20.700 --> 00:30:22.880
of the box. And with Kubernetes, you have to
00:30:22.880 --> 00:30:24.940
become like a pretty decent Kubernetes expert
00:30:24.940 --> 00:30:27.400
to do. Like the simplest one that I can give
00:30:27.400 --> 00:30:32.079
you is we do a scale to zero application deployment
00:30:32.079 --> 00:30:34.460
by default. I have a really good reason for doing
00:30:34.460 --> 00:30:35.779
that is because you're going to run this on a
00:30:35.779 --> 00:30:38.640
machine, your own computer somewhere. And if
00:30:38.640 --> 00:30:41.559
it's an app that only is used once every month,
00:30:41.740 --> 00:30:45.140
don't run it. Don't run it. Just let
00:30:45.140 --> 00:30:47.000
it run the one time a month that needs to run.
00:30:47.079 --> 00:30:49.480
Like don't take up resources running it,
00:30:49.559 --> 00:30:52.980
you know, for all 29 other days. And you can
00:30:52.980 --> 00:30:54.619
do that with Kubernetes. There's a bunch of Knative
00:30:54.619 --> 00:30:56.160
stuff. There's a bunch of pieces that you can
00:30:56.160 --> 00:30:58.079
sort of layer on top of Kubernetes in order to
00:30:58.079 --> 00:31:01.460
do that. But now you become an expert in a layer
00:31:01.460 --> 00:31:04.269
on top of Kubernetes instead of. It's just a
00:31:04.269 --> 00:31:06.109
thing that does the thing by default, right?
00:31:06.529 --> 00:31:09.670
And so it becomes a question of those trade -offs,
00:31:09.730 --> 00:31:12.289
right? Like we don't do as much as Kubernetes,
00:31:12.609 --> 00:31:15.690
obviously, but the things that we do do are designed
00:31:15.690 --> 00:31:20.130
specifically to do in that way. And we think
00:31:20.130 --> 00:31:21.710
that those are the right trade -offs to make
00:31:21.710 --> 00:31:27.049
for your teams of like one to 20 is really sort
00:31:27.049 --> 00:31:29.940
of the sweet spot that we target. So what kind
00:31:29.940 --> 00:31:32.859
of applications or workloads or problem statements
00:31:32.859 --> 00:31:36.740
do you see customers bringing to Miren? What
00:31:36.740 --> 00:31:38.640
do they come to Miren for that they wouldn't
00:31:38.640 --> 00:31:40.579
go to Kubernetes for? Yeah, I mean, they come
00:31:40.579 --> 00:31:44.819
to us a lot of times. The ambiguity around the
00:31:44.819 --> 00:31:47.460
Heroku situation is one reason people come to
00:31:47.460 --> 00:31:48.940
us. And they're like, I don't really want to
00:31:48.940 --> 00:31:52.420
take on, I really want a system that just does
00:31:52.420 --> 00:31:54.299
the system the right way. I don't want to become
00:31:54.299 --> 00:31:57.259
an expert in how to layer a deployment system
00:31:57.259 --> 00:32:00.589
on top of Kubernetes. That is like all all the
00:32:00.589 --> 00:32:02.190
helm charts and all the things and keeping up
00:32:02.190 --> 00:32:03.569
on all the versions and all that kind of stuff.
00:32:03.609 --> 00:32:05.609
They're like, I don't want that. I don't have
00:32:05.609 --> 00:32:08.630
the time or the team size to do that. But I want
00:32:08.630 --> 00:32:12.450
a thing that is that just just works. Like what
00:32:12.450 --> 00:32:14.410
we think about it, we think about ourselves as
00:32:14.410 --> 00:32:16.869
being application focused and workflow focused.
00:32:16.950 --> 00:32:18.630
So people come to us and they're like, hey, I
00:32:18.630 --> 00:32:20.349
just want to deploy my stuff. We're like, great.
00:32:20.410 --> 00:32:22.150
That's what we do. We deploy your stuff for you.
00:32:22.450 --> 00:32:26.170
You know, we we support. specifically because
00:32:26.170 --> 00:32:28.549
this is where 90 % of the workload is now. It's
00:32:28.549 --> 00:32:31.650
like we're supporting HTTP applications, you
00:32:31.650 --> 00:32:33.930
know, out of the box as the first class thing,
00:32:34.009 --> 00:32:36.190
because that's where people are, right? We have
00:32:36.190 --> 00:32:38.609
stuff for other, we sort of started to layer
00:32:38.609 --> 00:32:41.829
on things for non -HTTP stuff, but like the bulk
00:32:41.829 --> 00:32:44.609
of our focus has been on how do we run HTTP based
00:32:44.609 --> 00:32:48.269
apps really well and quickly. And so. Where do
00:32:48.269 -->
00:32:51.049
you think deployment and IaC tooling goes next?
00:32:51.150 --> 00:32:53.920
Especially. with AI speeding up code changes?
00:32:54.140 --> 00:32:56.980
Yeah, I mean, I think that people, one of the
00:32:56.980 --> 00:32:59.720
things people ask me is like, is Miren in danger
00:32:59.720 --> 00:33:03.059
of having an AI just code something to use AWS
00:33:03.059 --> 00:33:05.759
that works the same way? And I think the answer
00:33:05.759 --> 00:33:08.740
right now is no. But I think that even in the
00:33:08.740 --> 00:33:11.559
longer term, it's probably also no. And here's
00:33:11.559 --> 00:33:15.779
my take. You can have an AI go through and code
00:33:15.779 --> 00:33:18.299
you up a deployment platform in the same that
00:33:18.299 --> 00:33:20.700
feels similar to the way that Miren feels today
00:33:20.700 --> 00:33:23.839
on top of AWS. It could, you know, those AWS
00:33:23.839 --> 00:33:25.500
graphs, it's like all the different services
00:33:25.500 --> 00:33:27.200
wired together. It could do that. It could wire
00:33:27.200 --> 00:33:29.440
all those services together. Is it on call? Is
00:33:29.440 --> 00:33:31.880
the AI on call when one of those things goes
00:33:31.880 --> 00:33:35.619
wrong? What happens when the AI... is looking
00:33:35.619 --> 00:33:38.319
at it again after a month of it working and is
00:33:38.319 --> 00:33:40.500
now trying to figure out something is wrong and
00:33:40.500 --> 00:33:42.380
it's looking at it again. It's like, this doesn't
00:33:42.380 --> 00:33:44.880
make any sense. Who set this up? I'm actually
00:33:44.880 --> 00:33:46.440
going to delete all these things and rebuild
00:33:46.440 --> 00:33:48.180
it from scratch. You're like, oh my God, don't
00:33:48.180 --> 00:33:51.480
do that, right? And so there's not the discipline
00:33:51.480 --> 00:33:54.380
in order to do that. Now, again, I think people
00:33:54.380 --> 00:33:56.119
will be like, well, maybe the AIs will get that
00:33:56.119 --> 00:33:58.839
discipline. Yeah, I mean, then you're effectively
00:33:58.839 --> 00:34:02.000
hiring an SRE to just build that up. And the
00:34:02.000 --> 00:34:04.079
question is, is that a good use of your time
00:34:04.079 --> 00:34:07.650
and or money? right to to have an SRE spend all
00:34:07.650 --> 00:34:11.630
their time trying to balance AWS services you
00:34:11.630 --> 00:34:14.650
know on one hand uh for some people maybe it
00:34:14.650 --> 00:34:17.610
is uh i would i think our bet is that it's not
00:34:17.610 --> 00:34:20.170
you know like it's just not it it doesn't make
00:34:20.170 --> 00:34:22.989
sense in the the larger frame of things and so
00:34:23.670 --> 00:34:28.030
The other point about AI and IaC is that one
00:34:28.030 --> 00:34:30.010
of the things that we've seen is we've sort of
00:34:30.010 --> 00:34:32.550
pointed AI at our tooling and said, hey, go deploy
00:34:32.550 --> 00:34:35.210
this thing, is that our surface area is very
00:34:35.210 --> 00:34:38.190
small, but very obvious for deployment. And so
00:34:38.190 --> 00:34:41.190
the AI gets it right every single time, right?
00:34:41.269 --> 00:34:43.110
Because it's kind of nothing to get wrong. It's
00:34:43.110 --> 00:34:45.090
like, hey, I see that I can set config variables.
00:34:45.309 --> 00:34:46.849
I see that I can do deployments. I see that I
00:34:46.849 --> 00:34:48.530
can do rollbacks. I see that I can get status.
00:34:48.610 --> 00:34:51.849
Great. So the surface area the AI has to interact
00:34:51.849 --> 00:34:54.070
with is small enough that it just doesn't get
00:34:54.070 --> 00:34:55.909
it wrong. Right. Because it's like there's not
00:34:55.909 --> 00:34:58.550
a bunch of ambiguity about what is actually going
00:34:58.550 --> 00:35:01.090
on. And then, you know, the AI knows that like
00:35:01.090 --> 00:35:04.690
what doesn't know anything. But the API knows
00:35:04.690 --> 00:35:06.769
that behind all of that is the machinery that
00:35:06.769 --> 00:35:08.769
keeps all that stuff working in the shape that
00:35:08.769 --> 00:35:12.329
the API is made to work from. And so, I mean,
00:35:12.369 --> 00:35:14.360
I wonder like. what we're talking about here
00:35:14.360 --> 00:35:16.639
is abstraction levels. I don't think abstraction
00:35:16.639 --> 00:35:19.760
levels are going to disappear as AI gets better.
00:35:19.920 --> 00:35:22.340
It's going to build new abstraction levels. And
00:35:22.340 --> 00:35:23.860
then you're basically like, okay, great. That
00:35:23.860 --> 00:35:25.780
abstraction level is done. We're good. We're
00:35:25.780 --> 00:35:28.239
solid. Now I can go to the next one. And the
00:35:28.239 --> 00:35:32.380
question is, do you, and this was the case for
00:35:32.380 --> 00:35:35.659
Waypoint way back in the day. People build deployment
00:35:35.659 --> 00:35:39.019
systems at their companies very commonly. And
00:35:39.019 --> 00:35:40.900
the thing that we would always hear when we first
00:35:40.900 --> 00:35:44.360
was working at Waypoint was that, I got halfway
00:35:44.360 --> 00:35:47.000
done building this cool ass deployment system.
00:35:47.340 --> 00:35:50.699
And then my boss told me, hey, knock it off.
00:35:50.900 --> 00:35:53.280
Go back to the thing that actually we do as a
00:35:53.280 --> 00:35:55.699
company. And so then they'd have this half built
00:35:55.699 --> 00:35:58.559
deployment system that is just kind of a hunk
00:35:58.559 --> 00:36:00.360
of junk because it never got finished. And then
00:36:00.360 --> 00:36:02.920
they have to just suffer. with it for years.
00:36:03.199 --> 00:36:05.019
And they're like, how come the deployment system
00:36:05.019 --> 00:36:06.880
is so weird that you have to like stand on one
00:36:06.880 --> 00:36:08.920
leg and shake a thing above your head in order
00:36:08.920 --> 00:36:10.619
to make it work? And it's like, well, because
00:36:10.619 --> 00:36:12.579
none of that was supposed to be there. That
was
00:36:12.579 --> 00:36:14.820
all supposed to be paved over. But we never got
00:36:14.820 --> 00:36:17.179
enough time to finish it because we're a mortgage
00:36:17.179 --> 00:36:19.760
writing company. Right. So like building a deployment
00:36:19.760 --> 00:36:22.860
system never made sense. And so I think that
00:36:22.860 --> 00:36:26.519
that's still true when people are. applying AI
00:36:26.519 --> 00:36:29.420
to AWS resources. It's just like, you're, you
00:36:29.420 --> 00:36:31.260
just needed a thing to work. You didn't need
00:36:31.260 --> 00:36:33.880
any of those other pieces. We want to be, we
00:36:33.880 --> 00:36:38.019
want to basically as Miren be a person who has
00:36:38.019 --> 00:36:40.260
opinions. You want to basically buy our opinions.
00:36:40.420 --> 00:36:42.940
You want, we want, you want to outsource your
00:36:42.940 --> 00:36:44.980
opinions about how this is supposed to work to
00:36:44.980 --> 00:36:46.719
us. And then we'll just give you a thing that
00:36:46.719 --> 00:36:49.519
works. Oh, that makes sense. Yeah. I, I fell
00:36:49.519 --> 00:36:51.320
into the same trap. I think with Jenkins, like
00:36:51.320 --> 00:36:53.360
most people do. Oh yeah. Me too. You know, you
00:36:53.360 --> 00:36:55.400
didn't have time to. to fix it. And it was a
00:36:55.400 --> 00:36:57.079
house of cards. It just works. So you just left
00:36:57.079 --> 00:36:58.920
it alone. Right. Don't touch it. Yeah. Yeah.
00:36:58.960 --> 00:37:02.119
No better to, to yeah. Go to a company that can
00:37:02.119 --> 00:37:04.360
give you the opinions around the stack. You pay
00:37:04.360 --> 00:37:06.860
for that, pay for that. You pay for that service
00:37:06.860 --> 00:37:08.940
that they're going to maintain. Yeah. I mean,
00:37:08.960 --> 00:37:11.500
I have, I know people who, uh, who pay for AWS
00:37:11.500 --> 00:37:13.380
consultants because, and then they're there,
00:37:13.380 --> 00:37:15.480
that's the same thing. Right. And, and in some
00:37:15.480 --> 00:37:17.900
ways, uh, you know, they're, you're paying for
00:37:17.900 --> 00:37:20.260
that person's opinions. So. Okay, so wrapping
00:37:20.260 --> 00:37:23.340
up, what's one industry belief about DevOps or
00:37:23.340 --> 00:37:25.860
platform engineering that you think is just wrong?
00:37:26.179 --> 00:37:32.260
I think that people fall into the trap of, I
00:37:32.260 --> 00:37:34.699
guess it's not DevOps. I'll say what I was going
00:37:34.699 --> 00:37:37.119
to say, which is, yeah, I was going to say that
00:37:37.119 --> 00:37:39.820
I think that people fall into the trap of feeling
00:37:39.820 --> 00:37:42.369
like, oh, I can't. i shouldn't rewrite that i
00:37:42.369 --> 00:37:44.630
shouldn't do how about my own version of this
00:37:44.630 --> 00:37:46.769
that and the other thing and i think that um
00:37:46.769 --> 00:37:49.309
it's a double -edged sword of basically being
00:37:49.309 --> 00:37:52.849
like oh don't ever always use this one version
00:37:52.849 --> 00:37:56.070
of this thing um even though it does no most
00:37:56.070 --> 00:37:58.030
of it doesn't do what we want but this one little
00:37:58.030 --> 00:38:00.809
slice does right i think that people should be
00:38:00.809 --> 00:38:05.670
more uh willing to say I'm going to do the thing
00:38:05.670 --> 00:38:08.309
that works best for me, and I'm going to do 5
00:38:08.309 --> 00:38:10.050
% of what the other thing does, but it's going
00:38:10.050 --> 00:38:12.429
to be the 5 % that I want and now I can own.
00:38:12.630 --> 00:38:16.449
It's sort of like the industry did a lot of hand
00:38:16.449 --> 00:38:18.750
-wringing around dependencies about 10 years
00:38:18.750 --> 00:38:21.650
ago, about how big your node_modules should be
00:38:21.650 --> 00:38:23.469
and all that kind of stuff, right? I think we're
00:38:23.469 --> 00:38:24.989
running into it again now with supply chain.
00:38:25.309 --> 00:38:28.789
And so I think that having your own versions
00:38:28.789 --> 00:38:32.969
of the things that fit just what you need, it's
00:38:32.969 --> 00:38:36.019
way underrated. So yeah, no, that's fair. I agree
00:38:36.019 --> 00:38:39.639
with that. Evan, where can people find more about
00:38:39.639 --> 00:38:42.980
you? You have sOCIals? Yeah, yeah. You can find
00:38:42.980 --> 00:38:46.960
me on Bluesky. I'm evanphx.dev on Bluesky.
00:38:47.280 --> 00:38:49.059
We have a Miren Discord. If you want to hit
00:38:49.059 --> 00:38:51.039
that up, that's at miren.dev/discord.
00:38:51.420 --> 00:38:53.480
Yeah, that's probably the main two places he
00:38:53.480 --> 00:38:55.860
says. Awesome. Appreciate it. Thank you so much,
00:38:55.880 --> 00:38:57.579
Evan, for coming on. Really appreciate it. Absolutely.
00:38:57.679 --> 00:39:00.619
Thanks so much. All right. That was my conversation
00:39:00.619 --> 00:39:03.630
with Evan Phoenix from Miren. My biggest takeaway
00:39:03.630 --> 00:39:06.889
from this one is that deployment is painful because
00:39:06.889 --> 00:39:09.570
it sits at the intersection of everything else.
00:39:09.769 --> 00:39:13.010
It is not just run the app. It is build the app,
00:39:13.150 --> 00:39:16.670
package it, push it somewhere, wire up the config,
00:39:16.989 --> 00:39:21.389
expose the service, watch the logs, handle rollbacks,
00:39:21.489 --> 00:39:24.429
and make sure the next person can understand
00:39:24.429 --> 00:39:27.949
what happened. And every time we try to simplify
00:39:27.949 --> 00:39:31.039
that, we usually move the complexity somewhere
00:39:31.039 --> 00:39:34.099
else. Sometimes that is fine. Good abstraction
00:39:34.099 --> 00:39:37.440
hides the right things, but bad abstraction hides
00:39:37.440 --> 00:39:41.019
the thing until production is broken. And then
00:39:41.019 --> 00:39:44.400
suddenly, the team that didn't need to know Kubernetes
00:39:44.400 --> 00:39:48.199
has to understand pod eviction, node pressure,
00:39:48.679 --> 00:39:53.139
ingress behavior, registry auth, and whatever controller
00:39:53.139 --> 00:39:57.260
is angry three layers down. I liked Evan's point
00:39:57.260 --> 00:40:00.679
that small teams often do not need the biggest,
00:40:00.800 --> 00:40:03.699
most flexible platform. They need something with
00:40:03.699 --> 00:40:06.719
a smaller surface area that handles the boring
00:40:06.719 --> 00:40:09.599
parts well. That does not mean Kubernetes is
00:40:09.599 --> 00:40:13.199
bad. It does not mean Terraform is bad. It just
00:40:13.199 --> 00:40:17.420
means flexibility has a cost. Every knob is a
00:40:17.420 --> 00:40:20.960
decision. Every decision becomes something to
00:40:20.960 --> 00:40:28.269
support. And eventually, the platform becomes its own pile of work. The AI
00:40:28.269 --> 00:40:31.829
angle makes that even more interesting. AI can
00:40:31.829 --> 00:40:34.630
generate deployment code. It can wire together
00:40:34.630 --> 00:40:37.889
cloud services. It can probably build something
00:40:37.889 --> 00:40:42.690
that works for a while. But is AI on call? Does
00:40:42.690 --> 00:40:45.809
the team understand what it created? Does the
00:40:45.809 --> 00:40:49.130
next person know why those decisions were made?
00:40:49.690 --> 00:40:53.050
That is why good abstractions probably matter
00:40:53.050 --> 00:40:57.039
more, not less. A small, clear deployment surface
00:40:57.039 --> 00:41:01.079
is easier for humans to use and easier for AI
00:41:01.079 --> 00:41:05.239
to interact with safely. So the practical question
00:41:05.239 --> 00:41:08.699
is not always how do we build the most powerful
00:41:08.699 --> 00:41:12.039
platform? Sometimes it is how little platform
00:41:12.039 --> 00:41:15.460
can we get away with while still shipping safely?
00:41:15.760 --> 00:41:19.019
I'll have links to Evan, Miren, the Discord,
00:41:19.260 --> 00:41:22.099
and anything else we mentioned in the show notes.
00:41:22.760 --> 00:41:25.960
If you enjoyed this conversation, follow or subscribe
00:41:25.960 --> 00:41:28.619
to Ship It Weekly wherever you listen to podcasts.
00:41:28.900 --> 00:41:31.460
You can also find the show notes and links over
00:41:31.460 --> 00:41:34.980
on shipitweekly.fm. Thanks for listening, and
00:41:34.980 --> 00:41:36.599
I'll see you later this week.