hpr-knowledge-base/hpr_transcripts/hpr2934.txt

Episode: 2934
Title: HPR2934: Server Basics 106: Namespaces and containers
Source: https://hub.hackerpublicradio.org/ccdn.php?filename=/eps/hpr2934/hpr2934.mp3
Transcribed: 2025-10-24 13:32:06

---

It's Thursday the 31st of October 2019, and this is HBR Episode 2934 entitled,
Server Basics 106, Namespace and Containers. It's hosted by Tlatu and it's about 34 minutes long
and carries the clean flag. The summary is, Tlatu talks about,
the Unshare and LX Seekmans. This episode of HBR is brought to you by AnanasThost.com.
Get 15% discount on all shared hosting with the offer code HBR15. That's HBR15.
Better web hosting that's honest and fair at AnanasThost.com.
Everybody, this is Tlatu and you're listening to Hacker Public Radio. This is an entry into my
my series that you probably thought I had abandoned called Server Basics. I think this is
episode let's say 106 and it's going to lead into unsurprisingly number 107 and I mentioned that
in tandem here because this is not the end goal. This episode is not the end goal. Like my previous
series on OpenVPN, I expect there to be two distinct episodes on this topic. Today's the one
that you're listening to right now is background information, probably not all that useful in terms
of on-the-ground typing commands on the job sort of stuff but in the next episode we'll talk about
something that is quite the opposite. It's all about what you could expect to find yourself
doing if being paid to manage the thing that I'm going to be talking about and what is that
thing that we're going to be talking about? It's containers. As I was doing the initial five or
six episodes of Server Basics, someone pointed out to me that a lot of what I was talking about
really was very basic. So basic in fact that the likelihood of you getting started in a
Unix admin, cis admin job and finding yourself doing the things that I was talking about, like I
don't know manually setting up firewalls and things like that is actually relatively low. That's
just kind of not the environment that you find these days. And I think there's a truth there
that the modern Unix cis admin, unless they are very specifically called a Unix cis admin
and they're often their little office just maintaining machines that drive the infrastructure.
Generally speaking, that's not the job anymore. There aren't multi-user Unix systems sitting
around driving an entire ecosystem itself. People aren't logging in like they used to with a user
account and managing things from from their terminal. That's just not what people have set up anymore.
Some kind of fancy intranet setup and the user accounts are all maintained by the software.
There are no user accounts, whether there's some user accounts, but there are no general user
accounts on the server. People aren't logging in themselves to your Unix server checking their
mail with the mail command and listing files or sharing files with FTP or something like that.
This just doesn't happen. So the new model being software-based, it requires a different set of
skills from the person maintaining the server. So one of the major new skills out there right now
that you can expect to find when you go out to get paid as an admin for a large entity,
our containers. That's where everything is at now and I've kind of conveniently skipped over
virtual machines both in content for hacker, public radio and in real life. It was a thing when I
got started in tech and it's just not a thing anymore. This was back and I don't know, let's say
let's call it 2009. Virtual machines were driving the startup that I got started in.
Everything was running on a virtual machine and then as I continued I found that the same was true
everywhere. There were no physical machines that you would log into. Well there were physical
machines but there were more machines than we had physical machines when you logged into a
development environment because it was all just a virtual machine. Every entity was a little
virtual host on what you could consider more or less as an orchestrator of pretend machines
and the orchestrator there were lots of different softwares out there for an admin to sit down
at their computer and see how many virtual machines were running and what kind of resources
they were all taking into kind of scale each one as needed. But that's sort of a thing of the past.
You don't really find that anymore and of course I say that and so am I out there. Several people
out there are thinking well that's my day job so I don't know how you could possibly say you
don't see that anymore. I'm meaning that you don't see that so much as an emerging sort of going
forward people aren't turning to virtualization or at least to virtual machines in that sense and
rather they are turning to this concept called containers. Now I've covered containers before
very briefly and in a very rudimentary level because that's all I had at the time and it was
frankly still pretty new at the time. It was episode 1522 how to use Docker in Linux containers.
You're welcome to check that that episode out. It might give you a little bit of an indication
of what I'm talking about and why but I did kind of want to cover. I wanted to peel some of that
back and talk about what it means to use a container, how that happens and some of the basic
interfaces involved. And then in the next episode I'll be talking about one of the few systems
that you may actually encounter when you walk into a job and are told you'll be managing containers
for us or you'll be managing these processes that are being spun up by developers and you'll be
monitoring those or whatever you might do. So let's get started. Now I should also make one more
caveat before I do that. This has been part of this has also been covered by me on a recent
GNU World Order episode. That's GNU World Order.info and it was episode 1339. I talked about the
NS Inter and Unshare commands. I'm going to be doing that right now. I'm going to be repeating
really the contents of that episode so you can kind of probably skip ahead a little bit if you've
already heard that. If you have not heard that then this will be new information. But I wanted to
repeat it here because it is kind of vital to understanding the difference really between a virtual
machine and a container because there are related technologies happening here. It's the virtual
machine is definitely unique and distinctive in itself but there are a lot of similar things
going on behind the scenes such as KVM and little things like that. So let's talk about namespaces
on Linux. Namespaces on Linux are like namespaces everywhere else. They qualify something in context
a some small element in the in a larger context. So in other words we use namespaces in real life
all the time. Really simple one. If I say I really like the enterprise that could mean at least
two different things. One it could mean that I'm a really big fan of Star Trek or in another context
it could mean that I really enjoy using software designed for big businesses of I don't know
1000 plus people. Is 1000 a lot? I don't know. 10,000 plus people. I don't know how big businesses are
but those are two valid uses of the term enterprise at least in today's world.
So to differentiate what I mean in real life I might use what we could call a namespace and
and there's it's not official but but I could say something like oh I really like the enterprise
from Star Trek and that has served as a namespace because it's qualified what I mean by this
shared term and that's what namespaces do. They take shared resources or shared terms or something
that is is global in in possible nature and it specifies exactly what it's referring to or
exactly it's scope I guess would be a better way to think of it. What's the scope of of this thing?
Well here's a namespace to define that scope. You see it in programming languages as well you'll
see it in well for instance in Python if you do an import of something effectively you've created
a new namespace so that commands or rather functions that may exist elsewhere are namespaced
by the fact that you've just imported I don't know beautiful soup for as bs now anytime you use
bs dot something we know exactly what the something is that it's not it's not a something existing
in a different library or a different module it is the one that falls under the beautiful
soup namespace or the bs namespace in a literal sense so those are namespaces and we see them
now rather recently on Linux you can see this in action yourself with a command called unshare
unshare launches a process under a namespace other than its default so here's a here's a quick
exercise to see this in action fire up a terminal and issue the command pit of tcs h unless your
default shell is tcs h in which case do something else like bash pit of bash right now I'm not
running tcs h and if I do a pit of tcs h it returns nothing at all you should have tcs h installed
on your system so all you've done right now is you've you've proven to yourself that you're not
running some shell and again since I'm doing this with tcs h that's what I'll continue to use
if you're using something else swap the terms so we've confirmed that tcs h is not running but we
can fix that we can make it run and we're going to make it run in a different namespace so what
you can do is pseudo unshare dash dash fork space dash dash pit space dash dash mount dash
proc tcs h so that's pseudo unshare space dash dash fork space dash dash pit space dash
mount dash proc space and then the command that you want to run so tcs tcs h in my in my case
when that launches it dumps you into a new session which is tcs h
and it seems like nothing's happened well watch this do a pit of tcs h from this prompt
and it will return that it is pit one now if you're a long-time Linux user you'll think
to yourself that's atrocious pit one is always in it and you'd be right for instance if you open
up a separate terminal session somewhere on your system and do a ps space one it will return
to you that pit one is in it why are you getting two different responses to the same question on
your system well it's because tcs h is running in its own namespace you can kind of see this
in a sort of a tangible way if you go to if you do an ls of slash proc slash asterisk slash
in s you'll see a bunch of namespaces listed there by by pit it won't mean anything to you but
those are namespaces that that's where all of that information is being tracked so if you do
from this new session not the tcs h session that you've launched if you do from your your your
session that you've just done a ps space one in and gotten in it back you can do a pit of tcs h
and you'll get a completely different number 26814 or 27333 or something like that it won't be
ps one it won't be pit one in other words and that's because this session sees the the truth
of of this process that you have unchared it sees where this process was forked from or was was
is branching off of and you can see it like well you know very literal way by doing a ps space tree
pipe that to less search for tcs h and you'll see that yeah tcs tcs h is running it's a child
process of this other bash process which is in turn a child process of something else and you'll
see it all kind of lined up there in your in your ps tree but from within that namespace when you
query your the system as as to what processes your system is aware of
it's only aware of of of the tcs h process that that that has been launched because
within its own namespace that is the first process that's the only process right now that exists
in its own namespace the advantage here as you may imagine is that if there's some process
running in a namespace and you ask that process to do something to I don't know another
process like process 42 then as far as that namespace knows there is no such thing as pit 42
that doesn't exist all that exists right now is pit 1 now if you launch something in your tcs h
namespace then then there will be a new pit within that namespace but the point is that
within that namespace it has limited knowledge of the system it that it itself is running upon
pretty fancy stuff and this is the the basis for for for everything that we're going to be
talking about for the rest of this episode and certainly for the next episode the ability to
create processes on a system that are are really not fully self-aware as far as they know they are
the center of the universe they are pit 1 nothing else exists they're running on a computer that
is imaginary and yet everything that they're running on is being provided by by something completely
different so if that sounds familiar that's because it is it's the basis for containers it's
the basis for Docker it's the basis for Linux container project is for the basis for Kubernetes
and a bunch of these other sort of cloud technologies that you hear about so let's take
this a little bit farther and this is again like everything in this episode this this episode is
purely academic this is all stuff that you're not likely to find yourself doing if you're getting
hired as a as a entry level cloud admin this is not stuff that you're going to be doing probably
not even stuff you're going to be doing is assist admin this this is kind of
developery stuff that maybe you would be doing I imagine if you were coming up with some kind of
new Linux container technology something that you're you know you're basing your work on something
in a very raw state and developing it up from that but there is a tool that we may as we'll
talk about and it's it's called LXC which is the it's the Linux container project you can
read more about it over at Linux containers.org they provide this this base system
called or this base command called LXC you can install it if it's not installed
Slack where it comes with it pre-installed so I don't have to install it but you can install it
on Fedora for instance it would be pseudo DNF install LXC and then you would probably want
for instance LXC dash templates and maybe LXC dash doc for documentation but the templates
is kind of important as well on Debian and Ubuntu and so on I think you can just do apt install LXC
okay so the first thing that you have to do in order for this to really work for you just like you
did for Docker in episode 1522 or not just like but but rather similarly to what we have to do in
episode 1522 with Docker you have to create some imaginary network bridges so that you can talk
to your container because once again these containers that you're creating are going to exist in
this name space whereas as far as they know they're the only thing that exists so you need kind
of a some kind of link from your host system to your containerized operating environment you can
do that on a modern Linux with the IP command pseudo IP space link space add space VR0 space type
space bridge so you've just created a bridge a bridge interface that can connect your make
believe network interface on a container that you create to your real life one on your on your host
system so in order for you to talk to a network interface you need an IP address you can choose
one yourself I think in the Docker episode if I'm I haven't actually listened to that recently
but if I recall correctly we just did that in a physical router I think so here we can just do
pseudo space IP space adder that's ADDR as an address space add as in we're going to add a new IP
and then space I don't know one nine two dot one six eight dot one six eight dot one six eight
so local private IP address that I'm choosing I typically use that when I'm doing this sort of thing
because that's a lot of easy stuff to remember space dev space VR0 so we're assigning that IP address
to the device VR0 make sure that it's up and running with pseudo IP link set VR0 up
that's just the old if config up command but done through the IP command so now you've got you've
got a little pretend network bridge to well frankly a network that doesn't doesn't work that doesn't
exist yet but that's fine we'll we'll fix that in a moment so LXC is is a baseline technology that
you'll see being used in other places you'll you'll see it sort of either being mimicked by docker
or you'll see it being mimicked by this or that Kubernetes project so this is kind of the
generic way of dealing with Linux containers and I keep saying Linux containers and I guess I
should probably stop and say okay this is what we're doing now we're we're creating a container
and what is a container well a container is like a virtual machine so it is it is as if though you
had downloaded and installed virtual box or vert manager and then you downloaded an ISO to some
Linux distribution and then you told your virtual machine application your virtual box your
vert manager that you're going to now create a computer inside your computer and on that computer
you're going to run this distribution that you just downloaded and as we know by now I think
in a virtual machine the whole stack is virtualized the computer decides through the software
that you have a make-believe CPU and you have some make-believe RAM and you have some pretend hard
drives and with these make-believe pretend resources it installs an Linux distribution and now
you're running it and you can open up a little screen that shows you the desktop and and you can
interact with this computer just as if though it was real a real thing but it's it's just virtualized
this has been around forever at this point and even even before then there's been virtualization so
it's been around forever I think we're all pretty familiar with it a container is going to feel
pretty similar to that but the thing to understand is that it's using namespaces it's doing all
of this stuff through the magic of namespaces and control groups on Linux within Linux such that
when you spin up a virtual OS rather than virtualizing everything like the CPU and the RAM
and so on it's using your host resources and it's just running little processes in this containerized
environment in this namespace kind of cordoned off environment so that as far as it knows it is
its own computer but you and I will always know that it's actually a process running on our host
in its own little namespace okay so in order to do that with lxc we need to create a configuration
file and we can just call it my mycontainer.conf is fine a couple of different lines one will be the
lxc dot uts name which is going to name the the the the thing that we're creating so we're just kind
of human readable so we can do that equals hacker public radio lxc dot network dot type equals
v et h virtual ethernet lxc dot network dot flags equals up lxc dot network dot link equals
well the the the network that we want to be able to control this through is BR zero so that's what
we'll use is the network link lxc dot network dot h w a d dr you can just make something up so
equals some kind of MAC address so for a colon 23 colon or 23 colon 23 colon 23 colon 23
something like that lxc dot network dot ipv4 you can just give it any ip address you want so
equals 192.168.168.1 of course I said any but it needs to be an ip address that will you know
on the same subnet as your as BR zero and you can also do the same thing for ipv6 if that suits you
and that would be just lxc dot network dot ipv6 equals and then some made up ip address so you can
save that file most of this stuff is is completely arbitrary and if you're wondering how you would
ever be able to duplicate that process without me reading off random arbitrary values to you how
you would ever remember that don't worry there are examples of these configuration files provided
to you from lxc so if you look in user share doc slash lxc there should be an examples directory
in there and it'll have a bunch of different configurations available to you so you can just
copy those modify them as needed and you'll have a container do you have a container yet actually
yeah you do you can see you can kind of see the I mean you don't have an os in your container or
anything but you can see that the container exists and that it that it's aware that it exists so
you can do pseudo lxc dash execute and then space dash dash name basic dash dash RC file and then
the path to your my container dot cons I just put it in my home directory and then space slash
bin slash bash space dash dash log file space my container dot log
now you're essentially inside you're running a bash shell inside your container
so you can do slash user slash s bin slash ip because your container has no ip command itself so
you're using the host ip command user has been ip at her show you'll be able to see the the fake
internet environment or ethernet environment in the container it'll be aware more or less of
of itself and you can do a you name dash a v and it'll say Linux hacker public radio
and then whatever kernel it is aware of which should be your host your host kernel
okay so then you can do you exit to get out of your container and now you can install
an operating system inside the container and the way to do that probably is to use a template
I say probably because if you haven't installed these then they might not be on your system it kind
of depends on how your lxc has been packaged up but if you look in slash user slash share slash
lxc there should be a templates directory there and there should be a bunch of different
bunch of different templates that you could feasibly use so if you did an ls on slash user share
lxc templates you'd see things like I don't know probably lxc dash alpine that's a pretty
popular one because it's apparently really small footprint kind of specially intended for
for containerization there could be probably an lxc arch Linux maybe a busy box a syntax
debian fedora slackware a bunch of different ones to choose from I'm going to use slackware so I'm
going to do pseudo actually I won't use slackware because I'm already running slackware so
pseudo lxc dash create space dash dash name alpine sample no I'll do name yeah no I'll do name
alpine and then space dash dash template alpine if you poke around in a separate terminal you can see
what's really happening is probably somewhere in var slash var slash lib slash lxc slash alpine in
my case whatever you're using would be different slash root fs lxc create is setting the root
of your container to that directory so as far as it knows that's its root file system so I mean
it's it's generally a mistake to try to find the container in in containers if you're looking for
a literal container it's kind of your your bound for disappointment there's there's not really a
nice and tidy kind of like this is these this is the place where all my files and processes and
things live because that's just not the structure of containers containers as I've already demonstrated
with unshare they exist in other places that that in the host system you know in the proc directory
and so on so it's not all consolidated as the name container would suggest but if you were looking
I guess for the container in the container it would be right there it would be in your var directory
under the in in this case it would be var lib lxc alpine root fs root fs is as far as that name space
knows that's slash all right so now once you've got a minimal operating system environment installed
you can start your container pseudo lxc dash start space dash dash name alpine space dash dash
rc file space till the my container dot comt now unlike the unshare command I mean this is leveraging
unshare in a center and well this isn't in a center but it's leveraging all this stuff but
unlike the unshare command that does not put you into the container it simply starts the container
so in order to attach to the container you're technically somewhere there in the back end you're
using in a center pseudo lxc dash attach dash space space dash dash dash rather name space alpine
so now you are you're you're entering your container now with with lxc dash attach and then the name
of the container the human readable name so once you're inside your container you're kind of
well you're in a very very bare bone system there's not a whole lot you're going to be able to do
hanging around in this empty container what you probably will want to do is install software so
that your container is serve it can can serve a purpose can run commands because right now it's
just it's the barest install of linux possible you can do the you name dash a v again and get some
information about about what your environment sees it won't be all that exciting but but once it's
installed what once the OS is installed for for whatever value installed serves in a container then
you'll have tools to its normal to its normal sort of you'll have access to its normal tool set
so that you can install applications you can run I don't know a web server in the container or
a WordPress instance or a development environment whatever you wanted to run inside of that container
that's the point of containers okay so in the next episode in this series and server basics one
six or seven I will talk about a slightly more pragmatic implementation of what I've covered here
and that is a system by which to to manage containers in a larger system so obviously you can
spin up Docker images or or LXC images left and right yourself at home but in the real world when
you go out into the the big big systems that's not how it's going to happen there's going to be a
system where you can look at all the containers that are running on on on the host you can make sure
that they are that that the ones that aren't being used are being spun down and the ones that are
that are required are being spun up and so on so we'll talk about that in the next episode until then
thanks for listening to Hacker Public Radio you've been listening to Hacker Public Radio at Hacker
Public Radio dot org we are a community podcast network that release the shows every weekday
Monday through Friday today's show like all our shows was contributed by an HBR listener like
yourself if you ever thought of recording a podcast and click on our contributing to find out
how easy it really is Hacker Public Radio was founded by the digital dot pound and the
infonomicon computer club and it's part of the binary revolution at binrev.com if you have
comments on today's show please email the host directly leave a comment on the website or record
a follow-up episode yourself unless otherwise status today's show is released on the creative
comments attribution share a light 3.0 license