0
1
00:00:01,700 --> 00:00:03,020
Verification or is back.
1

2
00:00:03,020 --> 00:00:05,960
What is your opinion on the usage of supervisor D.
2

3
00:00:06,020 --> 00:00:08,810
So that's a great question.
3

4
00:00:08,870 --> 00:00:14,780
Supervisor D I'm assuming what you're referring to is putting supervisor in the container and running
4

5
00:00:14,780 --> 00:00:18,440
it as the PID one or the root process right in that container.
5

6
00:00:18,500 --> 00:00:21,100
That is fine if you need to do it right.
6

7
00:00:21,110 --> 00:00:25,910
I know Docker is best practices are never more than one app per container.
7

8
00:00:25,910 --> 00:00:32,750
And I like to expose sort of take that idea and say it's really never more than one concern per container.
8

9
00:00:32,750 --> 00:00:37,170
And when in some cases that concern might be multiple executable.
9

10
00:00:37,340 --> 00:00:40,640
And so you would need to start of those with something else.
10

11
00:00:40,640 --> 00:00:43,090
And it could be some sort of fancy shell scripting.
11

12
00:00:43,250 --> 00:00:45,220
I like using supervisor D.
12

13
00:00:45,230 --> 00:00:56,690
In fact if you look at my HP repo I do that I have a horribly outdated P HP repo for level so I definitely
13

14
00:00:56,690 --> 00:01:01,280
need some updates to use some of the modern new Docker features but it shows off a couple of things
14

15
00:01:01,520 --> 00:01:10,040
that are unique and one of those is that for PDP when you're using NPM you need the PDP executable running
15

16
00:01:10,100 --> 00:01:16,220
and the engine X executable running and you could run those in separate containers and that would probably
16

17
00:01:16,220 --> 00:01:17,830
be the official Docker way to do it.
17

18
00:01:17,840 --> 00:01:22,460
If you were following their guidelines but their guidelines are just that they're just like best practices.
18

19
00:01:22,460 --> 00:01:26,420
They're the place to start but it doesn't mean you have to stay there if you know what you're doing
19

20
00:01:26,420 --> 00:01:27,200
and why you're doing it.
20

21
00:01:27,590 --> 00:01:35,690
So in my case I would in this particular example I combine both engine X and the p HP FGM into the same
21

22
00:01:35,690 --> 00:01:43,040
container image and I run them both with supervisor D and so it adds another layer of complexity because
22

23
00:01:43,040 --> 00:01:47,450
now I have Docker running a container that's actually running supervisor as supervisors running engine
23

24
00:01:47,450 --> 00:01:51,670
X and FGM but I find that it works well and I've ran.
24

25
00:01:51,710 --> 00:01:56,910
I've run this on very large global Web sites that are running in lots of regions and abuse.
25

26
00:01:56,930 --> 00:02:04,030
So this design is definitely time tested and is used in heavily heavy scenarios.
26

27
00:02:04,040 --> 00:02:09,070
The reason I like this is because Docker and swarm out of the box don't have the concept of a pod.
27

28
00:02:09,170 --> 00:02:14,720
If you're in Cuba and 80s world you have a pod concept and in that case you could run the engine X and
28

29
00:02:14,780 --> 00:02:22,960
the PSP side by side in their own containers inside the same pod and that pod is deployed together.
29

30
00:02:22,970 --> 00:02:27,260
So when it's deployed onto a server you would know that you'd get both and then they would both be able
30

31
00:02:27,260 --> 00:02:31,790
to use the sockets if you're familiar with this stuff and PSP they use the sockets to talk back and
31

32
00:02:31,790 --> 00:02:37,740
forth and that's very efficient but in the docker swarm way you would do it of the box that way you
32

33
00:02:37,740 --> 00:02:40,480
would have to talk over TGP and that adds lots of lag.
33

34
00:02:40,520 --> 00:02:45,950
I mean that lots of lag it's maybe fine for some people but when you're talking about going across server
34

35
00:02:45,950 --> 00:02:51,500
instead of just talking over a local socket inside the same kernel there's definitely added you know
35

36
00:02:51,530 --> 00:02:55,600
possibly multiple milliseconds if not tens of milliseconds of delay there.
36

37
00:02:55,610 --> 00:02:59,650
So I like to run them both in the same container and I use supervisor D to do that.
37

38
00:02:59,810 --> 00:03:01,250
So check out that example.
38

39
00:03:01,280 --> 00:03:03,620
I will throw that in chat.
39

40
00:03:03,620 --> 00:03:05,960
I think I saw someone post that in there already.
40

41
00:03:07,250 --> 00:03:11,160
So hopefully that helps you out and that answers your question.
41

42
00:03:11,330 --> 00:03:15,050
There's only a few places by the way that I use supervisor D and that's one of them.
42

43
00:03:15,320 --> 00:03:20,720
The other one might be like cold fusion because it also needs like a patch or something on the front
43

44
00:03:20,720 --> 00:03:23,670
end in order to work or area Gen-X.
44

45
00:03:23,690 --> 00:03:28,670
So anytime you have something where you've got a back in binary that has to run to Amul to run your
45

46
00:03:28,670 --> 00:03:32,490
code and then you need some sort of web server in front of it and they have to both work together to
46

47
00:03:32,810 --> 00:03:37,550
to deliver the service it total because they're always going to be deployed together and one doesn't
47

48
00:03:37,550 --> 00:03:38,630
work without the other.
48

49
00:03:38,630 --> 00:03:44,690
It makes complete sense that they go together in that case in the same image especially when you scale
49

50
00:03:44,690 --> 00:03:50,600
them if you're always going to scale them one for one because a lot of things with containers you would
50

51
00:03:50,600 --> 00:03:53,900
want to scale differently right you're gonna scale the number of API is different you're going to scale
51

52
00:03:53,900 --> 00:03:55,210
the number database container.
52

53
00:03:55,220 --> 00:04:00,430
So in those cases you definitely with the flexibility and you want to build to update them on separately.
53

54
00:04:00,470 --> 00:04:05,090
In this case I'm just going to update the app at the same time and I'm always going to replace the containers
54

55
00:04:05,090 --> 00:04:08,930
at the same time so it makes a lot of sense to do those together so hopefully that helps.