[00:00.000 --> 00:13.000]  You're good to start.
[00:13.000 --> 00:18.000]  Okay, so hi everyone.
[00:18.000 --> 00:26.000]  My name is Mohamed and together with Mikoia we work both at Canonica and today we will be talking about
[00:26.000 --> 00:31.000]  on the road to managing databases, which is a canonical effort to make open source operator
[00:31.000 --> 00:41.000]  as we will see for a large, let's say, you know, landscape of open source databases that we will check later on.
[00:41.000 --> 00:51.000]  So in the presentation we'll start by explaining what we are aiming for, basically the vision of our, let's say, open source operator.
[00:51.000 --> 00:59.000]  We will go into a bit more detail with Mikoia about the database that we did choose and of course we will focus on MySQL for the story.
[00:59.000 --> 01:11.000]  We will explain where we stand right now, we will see a demo and then we will finish about, by speaking about the roadmap for in particular SQL databases.
[01:11.000 --> 01:14.000]  So what we are aiming for at Canonica now.
[01:14.000 --> 01:19.000]  So we have, let's say, our building block for operators.
[01:19.000 --> 01:28.000]  So Juju is our operator, by more, is the technology that allows us to build quite easily operator for pretty much anything.
[01:28.000 --> 01:33.000]  In this case, we are using it to build operators for databases.
[01:33.000 --> 01:40.000]  And what Juju allows us actually is to kind of abstract the, what we call the backing clouds, yeah.
[01:40.000 --> 01:47.000]  Meaning that it can work with pretty much, I mean, if there is a list that you can find,
[01:47.000 --> 01:54.000]  hopefully later on when we will share this slide in, let's say, an internet where we list the exact list of backing clouds
[01:54.000 --> 02:00.000]  and all of the ones that I mentioned here are supported and including orchestrators on top of backing clouds.
[02:00.000 --> 02:08.000]  So to build is one of the LSD which can allow you, for example, to emulate highly available setup in a single machine or even across machine.
[02:08.000 --> 02:14.000]  It's something that you can use as well as your backing cloud or infrastructure provided if you wish.
[02:14.000 --> 02:23.000]  And on top of that, we are building open source operator for a list of databases, so MySQL for sure.
[02:23.000 --> 02:29.000]  But Postgres, good to be, the 20 version, open source and more to come.
[02:29.000 --> 02:38.000]  And the idea is that we would like to automate a number of services or, let's say, you know, operation around databases,
[02:38.000 --> 02:44.000]  including employing a highly represented, as you will see later, scaling out and in, and ignoring node,
[02:44.000 --> 02:52.000]  scaling up and down, adding resources or removing resources, and of course doing backups and restores and upgrades.
[02:52.000 --> 03:01.000]  So this is basically our build-up, and Nicolas, I think, he will talk about the database.
[03:01.000 --> 03:22.000]  So unfortunately, Amazon has haven't killed all the bases, especially on premise, so we will try.
[03:22.000 --> 03:37.000]  Yeah, basically what we are building on which text tag we are building our solution for MySQL is for sure not MariaDB,
[03:37.000 --> 03:40.000]  or what reason we will use MariaDB.
[03:41.000 --> 03:48.000]  We are using MySQL, and we are using in a database cluster.
[03:48.000 --> 03:57.000]  Of course it is group replication, virtually synchronous replication of sort of synchronous replication.
[03:58.000 --> 04:10.000]  We are using MySQL Shell, because a lot of kind of knowledge of Oracle experts have been created,
[04:10.000 --> 04:19.000]  and it creates a lot of painful things for us, but we are suffering and continue using MySQL Shell.
[04:20.000 --> 04:30.000]  We are using MySQL Router, Percona Extra Backup, and for metrics node Percona Monitoring Tools, sorry,
[04:30.000 --> 04:33.000]  on the MySQL Extra.
[04:33.000 --> 04:40.000]  Yeah, so group replication, why group replication?
[04:40.000 --> 04:44.000]  Because that's simple, we don't want to lose data.
[04:44.000 --> 04:56.000]  And we want to have good performance in case of if you compare it to semi-synchronistic.
[04:56.000 --> 05:00.000]  Yes, it looks like that simple.
[05:00.000 --> 05:17.000]  In fact, yes, on the top of Juju, it is additional sub-layer in the system, but anyway it simplifies a lot of operations.
[05:18.000 --> 05:33.000]  Yeah, how it looks in reality, what we have right now, in that concrete moment,
[05:33.000 --> 05:44.000]  with simple Juju Deploy MySQL command, you can get automatically deployed cluster in a DB cluster,
[05:45.000 --> 06:00.000]  you can scale it, it has self-healing, meaning that if one member dies, it can be recolored on start,
[06:00.000 --> 06:09.000]  if data not lost, it can do incremental state transfer, it can do full state transfer,
[06:09.000 --> 06:13.000]  if you lost the full data or data not recolorable.
[06:13.000 --> 06:19.000]  It also can self-heal after full cluster crash.
[06:19.000 --> 06:25.000]  It is not fully automated thing in case of MySQL.
[06:25.000 --> 06:33.000]  I saw today news from Poland that it is automated already in Galera 4, but not on MySQL.
[06:33.000 --> 06:35.000]  Full cluster crash.
[06:35.000 --> 06:43.000]  Yeah, of course, vertical horizontal scaling in meaning that you can add members,
[06:43.000 --> 06:47.000]  and you can remove members from this old hardware.
[06:47.000 --> 06:55.000]  We have kind of user management in meaning operator,
[06:55.000 --> 07:03.000]  operator is application, use set of internal system users for backups,
[07:03.000 --> 07:07.000]  for monitoring, for administration of cluster.
[07:07.000 --> 07:13.000]  We have user management for system users, and for Juju native applications,
[07:13.000 --> 07:19.000]  which doesn't say anything for you, I'm sure, we also have user management,
[07:19.000 --> 07:29.000]  so when the new application connects to our Juju, it receives automatically MySQL user,
[07:29.000 --> 07:35.000]  and the user can rotate it, and the user has limited amount of operations.
[07:35.000 --> 07:39.000]  Yeah, we have MySQL router in place.
[07:39.000 --> 07:47.000]  We see it as, of course, thing which should run on application side,
[07:47.000 --> 07:55.000]  and it is designed in our stand to be run on application side, works well.
[07:55.000 --> 08:01.000]  If you have the same version between MySQL and MySQL router, if you are lucky,
[08:01.000 --> 08:07.000]  and can update your application servers, that is, database servers, of course.
[08:07.000 --> 08:15.000]  If that condition is true, MySQL router is a great product.
[08:15.000 --> 08:23.000]  Also, encryption in transit, and some basic certificate management,
[08:23.000 --> 08:31.000]  centralized certificate management, and, of course, with several,
[08:31.000 --> 08:37.000]  maybe we could go already, merge it back of functionality.
[08:37.000 --> 08:41.000]  Not resource yet, but we are working hard.
[08:41.000 --> 08:45.000]  That's probably about current state.
[08:45.000 --> 08:51.000]  So, how this magic works in reality, it looks like that.
[08:51.000 --> 08:55.000]  Juju is deploying the MySQL.
[08:55.000 --> 08:59.000]  It runs three members.
[08:59.000 --> 09:03.000]  If you see something, you are lucky.
[09:03.000 --> 09:09.000]  Active addresses.
[09:09.000 --> 09:13.000]  Yeah, it is repeating.
[09:13.000 --> 09:17.000]  Another, for example, another example of scaling.
[09:17.000 --> 09:23.000]  Again, if you are lucky and see something on the screen, Juju, you need one.
[09:23.000 --> 09:29.000]  You can see the stuff, and you need it.
[09:29.000 --> 09:33.000]  It is quickly because the database is empty.
[09:33.000 --> 09:39.000]  But, in fact, if you have a lot of data, when a new member comes,
[09:39.000 --> 09:45.000]  it is needed to make a transfer of data from old member to new member.
[09:45.000 --> 09:52.000]  Of course, there is in the DG cluster, so the functionality is automatically in the DG cluster.
[09:52.000 --> 09:58.000]  No worries, automatically, not always works in case of MySQL.
[09:58.000 --> 10:04.000]  You need, sometimes, to force SST, and we are doing that functionality now.
[10:04.000 --> 10:10.000]  Yes, Kenny, I'm making notes. It doesn't work.
[10:10.000 --> 10:14.000]  I can't see anything.
[10:14.000 --> 10:16.000]  I can show you a code.
[10:16.000 --> 10:24.000]  That's probably it about live demos and Roslap.
[10:24.000 --> 10:32.000]  What is it saying?
[10:32.000 --> 10:36.000]  I'm just making this from addition, so we will talk about what we are planning to do next.
[10:36.000 --> 10:40.000]  I hope we are back now.
[10:40.000 --> 10:44.000]  Okay, that is basically what we called the entity before.
[10:44.000 --> 10:50.000]  We are using, let's say, the naming speed there for the control.
[10:50.000 --> 10:54.000]  So, the current version that we will be doing, say, around April,
[10:54.000 --> 10:58.000]  will contain what you can see, of course, already.
[10:58.000 --> 11:04.000]  But we are, as we are working on phasing the resource, so as I said, MySQL is not yet merged,
[11:04.000 --> 11:06.000]  but we are working on it.
[11:06.000 --> 11:10.000]  So, we haven't done a documentation, but it's not so difficult yet, and we are working on it,
[11:10.000 --> 11:12.000]  making that available for you.
[11:12.000 --> 11:18.000]  Observability, so give this mention already that we are using, let's say, an exporter.
[11:18.000 --> 11:22.000]  Hopefully, you will have all the stats, okay, that is provided, actually, by Juju.
[11:22.000 --> 11:24.000]  Not all of the work is done.
[11:24.000 --> 11:28.000]  I mean, some of the work is already done by Juju, let's say, and for Shaxx and Beach.
[11:28.000 --> 11:34.000]  And, of course, the English and English team for our architecture, including the operator,
[11:34.000 --> 11:36.000]  will probably be snazzy.
[11:36.000 --> 11:40.000]  But we might be using it for some reason.
[11:40.000 --> 11:42.000]  So, this is basically where we are operating on.
[11:42.000 --> 11:46.000]  We have confidence that this should be delivered.
[11:46.000 --> 11:50.000]  Just to give you, let's say, an impression.
[12:04.000 --> 12:06.000]  Thank you.
[12:34.000 --> 12:36.000]  Thank you.
[13:04.000 --> 13:06.000]  Thank you.
[13:34.000 --> 13:36.000]  Thank you.
[14:04.000 --> 14:06.000]  Thank you.
[14:34.000 --> 14:38.000]  Take a look.
[14:38.000 --> 14:42.000]  So, that's it about our presentation.
[14:42.000 --> 14:48.000]  I brought some sweets as well for you that you can enjoy for free.
[14:48.000 --> 14:52.000]  And that would be even happier if you ask questions.
[14:52.000 --> 14:56.000]  May I ask a question about...
[14:56.000 --> 15:02.000]  Obviously, I think you guys raised these happy words.
[15:02.000 --> 15:06.000]  Group reputation has a maximum number of members.
[15:06.000 --> 15:12.000]  So, if somebody wants more members, more feedback, that is your system.
[15:12.000 --> 15:16.000]  Yes, so, basically, that's right.
[15:16.000 --> 15:24.000]  Juju group reputation has limitation of 9.11 members.
[15:24.000 --> 15:26.000]  What is suggested way?
[15:26.000 --> 15:30.000]  We think that we will use cluster sets.
[15:30.000 --> 15:36.000]  Very soon, we have a design how to make this happen in case of Juju.
[15:36.000 --> 15:44.000]  So, basically, we will have several energy being cluster sets, if you know what it is.
[15:44.000 --> 15:48.000]  Basically, it is two different group reputation clusters.
[15:48.000 --> 15:52.000]  These are synchronous replication between them.
[15:52.000 --> 15:56.000]  One, two, three, five.
[15:56.000 --> 16:02.000]  And is this underlying CR exposed to the user?
[16:02.000 --> 16:06.000]  Or do they have to do CLI?
[16:06.000 --> 16:10.000]  Dependent on what?
[16:10.000 --> 16:20.000]  So, if you use teachers of manager database, I mean, you want to scale, you want to enable DLS,
[16:20.000 --> 16:24.000]  you want to change the class for administrative system user.
[16:24.000 --> 16:30.000]  In that case, you are using Juju CLI.
[16:30.000 --> 16:36.000]  If you want doing something with database, you are creating database.
[16:36.000 --> 16:40.000]  But it depends.
[16:40.000 --> 16:42.000]  Just to explain this.
[16:42.000 --> 16:48.000]  What you can say is that, I mean, the operator, our code is really about managing database.
[16:48.000 --> 16:52.000]  You can get the regular secret connection and do whatever you would like.
[16:52.000 --> 16:56.000]  Your application does not have to be within Juju.
[16:56.000 --> 17:02.000]  We can configure it in a way so that your application is outside.
[17:02.000 --> 17:06.000]  At least, the part I want you to see is that it is a part of Juju's operator.
[17:06.000 --> 17:10.000]  You can install a CRV that exposed the CRs.
[17:10.000 --> 17:12.000]  No, no, no.
[17:12.000 --> 17:22.000]  It is, yes and no, Juju is operator in meaning of what is operator, but it doesn't work to see it is.
[17:22.000 --> 17:34.000]  You can think of it as an orchestrator that can work with not only, I mean, Kubernetes as the one of multiple players.
[17:34.000 --> 17:40.000]  But it's a good way of thinking about it as an orchestrator.
[17:40.000 --> 17:42.000]  It depends the character.
[17:42.000 --> 17:50.000]  You can place a role as an intermediate layer.
[17:50.000 --> 17:56.000]  That is, of course, completely my base on the slide.
[17:56.000 --> 18:06.000]  And at the same time, as a different user, I would guess that it plays the role of an intermediate layer between the user and the system.
[18:06.000 --> 18:12.000]  Because, well, I see that all the interaction was there.
[18:12.000 --> 18:18.000]  So, is it true that it's actually a component which plays on both fields and funds?
[18:18.000 --> 18:20.000]  Yes, let me answer.
[18:20.000 --> 18:34.000]  So, basically, Juju is kind of automation, a middle layer automation, but we built our operators on top of Juju.
[18:34.000 --> 18:38.000]  In that case, we don't care what the file system.
[18:38.000 --> 18:44.000]  We say, hey, give me the disk there, into that directory.
[18:44.000 --> 18:48.000]  And don't care about multiple field networking and so on.
[18:48.000 --> 19:00.000]  And it has a lot of patterns for process management, which you can say, hey, run me that process.
[19:00.000 --> 19:08.000]  But on that privilege, constantly, or run it on sports when I am doing some action.
[19:08.000 --> 19:21.000]  So, it is kind of operator framework, but operator framework, which uses some unified level of hardware under the hood.
[19:21.000 --> 19:27.000]  And it equally works well on open stack as on Kubernetes.
[19:27.000 --> 19:32.000]  So, it is a bit more advanced, I would say.
[19:32.000 --> 19:37.000]  And not the same plan, it corresponds to the interaction with the user.
[19:37.000 --> 19:39.000]  Yes, absolutely.
[19:39.000 --> 19:43.000]  Thank you.
[19:43.000 --> 19:45.000]  Perfect.
[19:45.000 --> 19:49.000]  One more question.
[19:49.000 --> 19:51.000]  Thank you.