GitLab Support Team, 21 May 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Installing Gitlab using Helm (GKE)

Description

We provision a K8s cluster in GKE and then set up a static IP and a DNS entry to eventually install Gitlab on said cluster, with Helm

A

Okay, I've started the recording um okay, so what we're gonna do today is we're gonna provision, a gke cluster uh kubernetes cluster and we're going to install going to install gitlab on uh on that cluster, and, let's see how I will share my screen.

B

A

Cool, let me know when you see it: yeah yeah, yeah, okay, so the first thing I want to do is go to the google ui and actually create the um kubernetes cluster.

A

That would be under kubernetes engine clusters, and I don't know how how much experience you have with this hey. Chris looks like he joined us. After all,.

B

ah Yeah welcome chris.

A

Thank you. Thank you, okay, um so I'm not gonna. Do the autopilot thing I I saw that that's something they've added recently, I'm just gonna figure, but I think autopilot also also works. It's just.

B

That I've actually not personally tried it today, so it.

A

B

A

Much freedom, it comes with some already configured parameters. So, let's put a name.

A

Let us test let's test, I guess that's one, that's one! That's fine! Okay, and let's do something in europe or better. Okay, let's see that um yeah, let's go with release channel. I think that.

B

A

They upgrade your your version whenever.

B

A

Available and yeah that should be fine. um I usually also name my resources pool okay, number of nodes.

A

Everything looks good, let's choose the nodes I'll do I guess. Let's try this. uh I think it.

B

Actually and yeah yeah we need, I I it usually has a problem if you provision these pages and all that, so I think the best thing is you go to the end and on something yeah and here yeah, the ones yeah these ones and one the first generation yeah.

A

uh I don't have any one: oh okay,.

B

And then we can do from there maybe standard four yeah. I I just do eight gigs of memory, because I yeah.

A

It works because of the.

B

Because of the provision and sometimes uh you'll find that some pods fail so yeah yeah.

A

Yeah, actually, I've had this also good.

B

Yeah, um so that.

A

Works good, um I think that's mostly it yeah.

A

Okay, uh I'm not exactly sure where this goes. uh I know there was like metadata to add, for example, frugal time for goalscape.

B

Yeah, that's the skip yeah, I'm.

A

B

A

This is the place to put it, but, let's see sure and then just create. I guess.

A

And while that is running it's here- and you can also see it here as a status- let's create a an external ip, because we're going to need that we're going to use the static ip version of the install. So we need to create an external ip reserve.

B

A

We're going to call it blood test.

A

Something one I think that's about it, let's put on europe as well, your request, that's fine reserved, okay, and with this this ip, uh let me see if it's actually been provisioned.

A

Okay, there it we're, is define at the end, so.

B

A

B

So you can just keep that the ip address we need it. Yeah.

A

Yeah just grab it yeah and yeah. Let's see where is it services cloud dns.

A

And then support resources and then just create a record.

A

I guess blood s test, one which will be my dns entry and then the ip is the one that I've copied from the external static ip and that should be it.

A

C

A

A

Let's actually look for it, maybe it's in a different page, yeah jump towards the next page, so there it is.

A

uh And let's see yeah, this is still ongoing, so let's wait for it to to complete. In the meantime, um this is the page uh I think most of us should use to. uh Is it.

B

On the wrong page,.

A

Yeah, it's it's not.

A

This is something I've used for elasticsearch, apparently.

A

C

Actually, this is.

A

Yeah, this is the one that I meant wrong.

B

A

um Yeah, so it's chart installation deployments, blah blah blah and it's uh install or something it's all the way at the end, yeah yeah, this one, oh yeah, I didn't mention the the prereqs that uh one would need for for this install, so you need to have cube ctl installed, uh helm as well as whether as g util, which is the google cloud uh command line utility.

A

It's not really necessary, but I'm gonna use it uh because I'm lazy and I don't wanna, add my configuration to cube ctl like the configuration of the new kubernetes cluster to qctl manually, or it's just going to give me a command.

A

Let's see in the meantime, if it's done yep, it's not going there. So I need to go manually uh and I'm going to show you what I mean by that um if you go here to connect, so this is the cluster that we've created in kubernetes and if you go to connect, there's this this line here that you can.

B

A

And it uses gcloud, and I'm just gonna do that, and this automatically adds the new kubernetes cluster to my cube, ctl config. So right now, if I do uh by the way uh k is an alias for cube. Cdl ctl.

A

Whatever so, this actually runs on the um the new cluster new kubernetes cluster, okay, so let's install gitlab now um and yeah, as I've said, I've already installed helm and I've also added the repo. So an additional step for those who are just starting up and just installed helm. They would also need to add the the repo chart repo and then maybe do a helm repo update. Actually I can do this as well.

C

A

Just do that now: okay,.

A

Okay, uh let's see do I have anything underneath no okay boom, it's it's uh it's clean! So let's grab this command like that and yeah.

A

It's gonna put it here without the update stuff, so uh timeout, it's explained also in the docs, but this means that every component will need to install uh in under 600 seconds. If, if that one single component goes over, the the command will fail. So it's not the entire flow, not the entire gitlab, but the the various resources that are spun up need to be under 600 seconds and then here we're gonna have to put in the the stuff that we've created so the domain and all that other stuff. So, let's see.

A

I guess I could grab them from the dns entry.

C

A

Both of them so the dns and.

A

The quite a few here, but this.

C

Is the one that I eat you forgot to put star.

A

C

You're right yeah because it will fail for pages.

A

C

A

Yeah there needs to be a star here because it's going to be github vlades.

A

Blah blah blah you'll see that in the in the ingress.

A

Okay, uh did I grab the right one? Maybe not. Let's try that again. What is it here? So, let's replace example.com with this okay.

A

Like that, and then the ip as well.

A

And then it will be your your email address, so mine.

A

This is for the ssl and I think, there's other stuff that we might want to explore. Like I don't know, the version.

B

For example, yeah we can use version, you can also use. Probably we can also think about. If you want to enable pages, graphene and all those things they are not, but yeah.

A

By default yeah, maybe I don't want to confuse anyone uh too much. I just want to give them the the bare minimum. The.

B

Default yeah yeah.

A

B

A

The bare minimum but yeah, uh let's just get git lab to work and then we can. We can do other stuff afterwards, um so version mappings.

A

So here is the where you get the actual version that you would need to pass to this command to get the actual version of gitlab. um So let's say you want 1311 to or let's do 1310.2, for example, which is quite popular. um We would just need to do this. um I'm just gonna do it here, so I don't do any extra spacing or whatever version like that, uh maybe equal actually and then maybe a good idea would be to create a namespace.

A

um You could install it to default, but it's I think it's a good practice to to do it under um under a namespace, especially considering um when you're trying to I mean if, if your install fails or you fail to uninstall it, um you can just delete the namespace and most of the stuff will be deleted. So that's kind of what I why I do a namespace. So, let's.

B

Actually, maybe just to add on that, the namespace is important, because even the the way the pods will be named, they will be named with that specific name space. So yeah.

A

You're right, yeah, yeah.

B

Yeah yeah actually.

C

The recommendation.

A

B

Yeah, okay, actually use maybe grit club or something I think it's the default in the documentation. Yeah.

A

B

A

So you can see it's it's been created one second ago, so we can create. We can pass this as well into the command. So, let's, let's do that for namespace and.

A

Then, okay- and I mean this- could be uh just as well helm install uh it doesn't have to be upgrade minus install. uh I don't know why it's put that way in the docks, but I guess this. This will also work when you change the version, for example,.

B

Yeah, you can, you can use it both to to upgrade so you can never install.

A

Yeah exactly you can that's why I'm putting an older version, so uh we can, I mean we can hopefully do earlier version, but still so let me double check. Does everything look right, um yeah, we'll see.

A

A

Good yep, those are normal, doesn't mean they're errors. I think.

B

A

Sylvester is actually working on removing some of those.

B

Yeah yeah, I am watching, we are almost done with the most of them yeah yeah, but.

A

They're they're reassuring because I know right now we're not stuck because I keep seeing the uh things.

B

That are going on uh no yeah, okay, it's not a fair review on uh 1.22, so yeah.

A

B

Yeah still safe.

A

B

You you upgrade the cluster to 1.22. Then you'll have a small problem. There.

A

Okay, so it looks.

B

Like it was successful,.

A

B

See so maybe just one last part we missed. We need to add this privilege for the runner, because it will not install the runner.

A

B

It will have a problem, so you need to add it as a privileged yeah. That last surprise.

A

uh Yeah like this, this.

B

One down there down there um there's a problem with privilege: yeah there's a one in there, it's like a notice or something yeah, but it still works anyway. So.

A

Yeah exactly, um I think, that's also in the documentation.

B

Yeah there's a mention of that when you want to ensure at least you have access to the runner yeah.

A

Anyway, uh okay, so now, if I do qct I'll get.

C

A

Okay, a bit of things.

B

So there should be our url yeah that one yeah yeah, but I it's still creating yeah.

A

Yeah that way, I I didn't recognize those so oh.

B

A

uh What they were.

B

Yeah they are the ones that I think it's the load, balancer that is being generated, they'll actually disappear after a while or something so. We have.

A

B

Ip, so it should be okay, yeah.

A

And let's see yeah you're right c, terminating.

B

Yeah yeah, the the good.

A

A

C

A

Guitar is running most, it says zero out of one.

A

Radius, huh you.

C

B

Yeah because there are some jobs that uh in in in it, which.

A

The web service is oh.

B

The web service and that's key okay, let's see I knew maybe we can describe the portal just see.

B

Maybe it's about the the resources we we did. Maybe it's not.

A

B

Yeah there's a possibility. ah We did the two cpus and you described that we just see. Probably it's an issue of resource.

A

Yeah, I I think that the one that I uh picked out initially e, that that didn't uh show me any any problems. But let's see.

A

A

Let's grab this one.

A

Let's see what it says.

B

It's okay actually.

A

Successfully pulled.

A

Doesn't look like it's.

B

Yeah, it looks like it's running, maybe yeah. It's actually running.

A

B

It is so bad watch.

A

The containers, so this is the init container part.

A

Okay, let's try it again.

B

From this circle,.

A

B

A

Fails momentarily but those are all getting retried. So, um okay, so this looks good. Let's see again the ingress stuff.

A

And let's grab this because we want to log into gitlab right.

A

And there's something in here: okay uh about.

A

To password the password, it's a secret yeah cube, ctl, yeah, blah blah blah so initial root. Let's get.

A

That um just secrets -10.

A

And then grab for.

A

This one, let me grab that whole.

A

A

Okay- and we change it with this one that should give us.

B

You need to go to.

A

C

B

As in the the namespace.

C

A

Yeah right sorry.

C

uh These are that that's strong, just.

C

A

C

B

uh There's something you did there that I'm not quite sure the the echo and.

B

Let's confirm maybe.

B

Yeah, that's different, actually yeah, so the person is.

B

A

Good cool so yeah. This means that we have a running kubernetes cluster on. Actually, let's see that lmls, uh minus and git lab uh yeah status deployed chart version, 410, 0 and app version 1310 too, um and actually we can do the same thing, but we can change the version.

B

Yeah we can upgrade the actually.

A

B

Also downgrade that's the thing. That's.

A

That's uh something that's uh that didn't occur to me, but uh it's relatively easy with helm. Let's say we do 497, that's that should work.

A

Take a little bit.

A

497 was what version ah 1397 your corresponds.

A

That's taking quite a bit.

C

A

Because it has to.

C

A

C

Spring get the data yeah.

A

Yep we can see that there's things um and if we do helm unless.

A

A

uh And if we do.

A

uh History, easter.

B

Than the git club.

A

Gitlab, that was the name of the.

B

A

Yeah we can see that we were first on thirteen ten two and now thirteen nine seven. uh I think we can also do a roll back hell.

B

A

B

Draw a roll back roll back that uh that.

A

B

There you can say it too.

A

Yeah you don't have to I mean I have one latest, you don't have to mention it. It's actually one uh the revision that we're going to minus, and here let's see.

A

I didn't check, probably the other resources were not ready yet.

B

But so in in case you have many deployments, that's why I was saying you can actually indicate the the the division number at the end. So.

A

You can uh what sylvester is saying is that here you can actually do well. Actually, let's do history. First, you will get one two three. Maybe I don't know it can continue for for a while uh and if you, if you want to pick a specific version that you want to roll back to it's this command, but then you just add the revision, one or three or two or whatever.

A

If you just do it like like this, it's to the latest, so it would be, it would be just the one before the current one um and you can actually see the status as well. So this is deployed and the other ones are superseded.

A

Okay, um yeah, I think that's that's kind of what I wanted to show on the recording. Is there anything else that might be interesting.

B

Now, just to mention, depending on your cluster, these warnings may appear or may not appear the current. We are on uh 1.19, so.

A

B

You are on a alessa version of kubernetes. You may not see these warnings that you are seeing here, because the departure- yes, yes, yeah yeah, so so.

A

B

A

It yeah and yeah also, if you just want to install the latest uh version, you just need to exclude it just do this, and this will install the newest gitlab version, provided you've updated the the repos initially so this command right here. Where is it this one.

B

B

Maybe one last thing: what if I want to clean up, how do I ensure that right.

A

Right right and actually I've tried this a few times. uh I I just pressed enter on the newest version, uh one minute um but yeah, that's simple, that's helen delete and then the name we'll see that in a second.

A

Is this the version where we're not going to see those warnings, because.

B

No, no, I I meant the cluster the you see. If you look, uh we we took the 1.19.

B

C

You went, you mean the version.

B

Of kubernetes, yes, yes, if you have like uh 17 or 16, you may not see some of this one yeah yeah, I'm just mentioning, because you may have a cluster that so it's just good to be aware that.

A

B

Will also disappear. Some of them actually are in the latest studies. So.

A

So right now we're on 13 11 4, which is, I guess, the latest.

A

And we were saying actually let me just do.

A

A

Now they're still doing work yeah, I just want to leave them uh come up, uh so we don't delete it uh while it's still uh bringing stuff up. uh Although.

C

It it should work yeah they should. This should be. Let's, let's do that.

A

um Delete, um that's it yeah, no other things necessary. Oh minus n,.

B

A

And yeah: let's see.

B

Yeah, uh maybe a quick note on this delete. It will not actually remove everything so.

A

I was going to showcase that.

B

A

True so, let's see bugs gitlab. That would be there's, obviously a bunch of stuff in terminating, but uh it's it's not gonna delete all the resources like, for example, let's see secrets.

A

So those were not getting terminated for users. I think um there's there's quite a few resources that will not get removed, um I'm not sure about pvcs yeah.

B

Even volumes yeah because, uh like now you view an ins you've deleted install another instance. You should be able to get your data there. So.

A

B

To explicitly remove the I think in the documentation, there is a mention of a complete cleanup of how you would go about doing a complete cleanup, yeah.

B

Delete resources or something yeah, so that's that's also good to mention yeah.

A

Yeah- it's not in this page but yeah, and I'm I'm suspecting that I have a ticket with someone who's trying to install runner, but it's getting into some weird errors that might be related to any proper cleanup uh that happened beforehand. So yeah yep. Okay, I guess that's that's about it! For for now, I'm gonna stop the recording uh feel free to reach out to us. If you have questions about what you've seen here, bye.

B