0:10 [Music]
0:12 Well ladies and gentlemen for giving us
0:14 the download on mastering aerospike with
0:16 Kubernetes operator. I invite now on
0:19 stage Miss Shri Nidhi Kashik principal consultant
0:21 consultant
0:30 So today's topic is for folks who have
0:33 already on aerospike or looking forward
0:36 to using cubernetes to manage aerospike
0:38 successfully. So this is the uh this is
0:40 all about that particular
0:43 topic. So first thing is a quick agenda
0:44 of what we are going to follow through.
0:46 This is a quick agenda. I'll just not go
0:48 through this for from a timing
0:52 standpoint. So what we have right now if
0:55 you're facing any of these issues or any
0:58 concerns related to running your
1:01 uh aerospect clusters manually unable to
1:05 deploy on time finding scaling issues
1:07 the amount of effort that you're uh
1:10 investing in upgrading is huge or you're
1:13 running or you're dependent on runbooks
1:15 provided either by yourself or developed
1:18 by aerospike or you're seeing something
1:20 like your Kubernetes this cluster is
1:21 being managed by centralized
1:24 decentralized team which in turn uh
1:26 causes something like low engineer
1:29 morale or non-adherance to standards. If
1:31 any of these is what you are facing
1:33 right now as of today here's this
1:35 subsequent slides which talks about how
1:38 we can elevate this problem. Now from a
1:40 consequence standpoint we have seen that
1:43 whenever these issues are being faced we
1:45 this leads to something like missed SLAs
1:47 increased cost in managing your home
1:49 ground op automation as well as missed
1:51 revenue opportunities because
1:53 you're investing more time in doing
1:55 mundane teams like upgrading or
1:57 maintaining your aerospike which should
2:01 be given out for more innovative uh
2:04 efforts. Now this brings up to the next
2:06 slide. What is what is AKO all about? So
2:10 AKO or Aerospike K operator helps you to
2:13 automate the deployment management of
2:15 Aerospike enterprise clusters on
2:19 Kubernetes. Then why Aerospike AKO?
2:21 Because this helps you to reduce the
2:24 resource consumption chances of error in
2:26 while you're managing these parameters
2:28 and make sure makes your S sur lives
2:30 easier. That is what AKO is here to help
2:33 you out with. Now this is a very simple
2:36 view of how it architecture looks like
2:38 wherein on your right hand side depicts
2:41 the user. User provides the inputs to
2:44 the CRD files and we have host of CRD
2:47 files in our repository which caters to
2:49 each and every use case that you have in
2:51 mind. Right? The example that we are
2:53 going to showcase is just for demo
2:55 standpoint but much more complex
2:57 requirements can be managed
3:00 and heavier aerospike instance can be
3:04 set up with minimal fuss. Now from a CRD
3:06 standpoint is more of an YAML file that
3:07 we have which calls out the
3:10 configurations associated with your uh
3:12 aerospike environment and we have this
3:14 Kubernetes aerospike operator which
3:17 interacts with your Kubernetes API and
3:20 it helps you to do the next activity
3:22 which is to help you to horizontally
3:25 scale vertically scale upgrade your
3:28 aerospike instances with no uh
3:30 interference from the user as well as to
3:33 try changing parameters also. right
3:35 which needs rolling restart of your
3:37 environment and all of this is happening
3:39 mind you without any sort of downtime
3:43 right that's the beauty of using AKO now
3:47 we are compatible on with Amazon EKS AKS
3:49 Azure Kubernetes services Google
3:50 Kubernetes services and if you have an
3:53 inos kubernetes setup we are compatible
3:55 with those as well the same yl file the
3:57 same yard stick holds good except the
3:59 parameters would change based on the
4:01 cubernetes in uh cubernetes setup that
4:03 you have at your
4:06 Now this is just a small example of how
4:08 uh configuration looks like and how
4:10 complex an aerospike environment would
4:11 look over here on your right hand side.
4:13 So we have an operator connecting to
4:16 your Kubernetes API. You have a option
4:19 to c capture your configuration maps.
4:20 Secrets like TLS files, your
4:22 configuration files can be kept under
4:25 this secret directory. In addition to
4:28 it, we can also have a aerospike
4:30 Prometheus exporter as a sidecar which
4:32 will help you to monitor these pods or
4:34 clusters effort effortlessly. And mind
4:36 you, this is a single pane of operating
4:38 window. We don't have to have multiple
4:40 operators. One operator manages the
4:42 entire fleet of clusters that you have
4:44 from a aerospike standpoint. That's the
4:45 beauty of
4:50 AKO. Now going to the demo. So the
4:54 intent of this particular demo is to
4:56 yeah let me let me give you a voice over
4:58 makes it much more interesting. So what
4:59 we are trying to do over here in this
5:01 particular case is we have set up a GKE
5:05 cluster on uh Google cloud. So on your
5:06 left hand window what we have done is as
5:08 you can see we have installed u uh
5:10 aerospec operator and this is a jump
5:12 server on which we have aerospec
5:14 installed and what we are trying to do
5:16 in this particular case is
5:19 to there's a sample of uh the repository
5:20 details that I was talking to you about
5:23 the github link which gives you these
5:25 particular parameters for example what
5:27 I've done is I've taken a yl file called
5:31 submit 6x6.7x wherein I am intending to
5:33 run something like set up a cluster
5:36 called a cluster with three nodes.
5:37 There's a three pods that you see under
5:39 the size and what version we are talking about
5:40 about
5:43 6.1.0.1 as a version and this also
5:46 contains a snippet of the storage class
5:48 that we use from a role-based access
5:50 control. We have the users that has to
5:53 be generated automatically and finally
5:55 the aerospect configuration on what sort
5:58 of in installation we want. In our case
5:59 what we have done is we have used
6:01 something called memory size a pure
6:02 inmemory cluster for just for an
6:04 example. Now what I've done is I've just
6:06 run the cubectl appli minus f command.
6:09 And if you can see here it says cluster
6:11 created. Please concentrate on your
6:13 right hand window. What you have done is
6:15 we have just tried to access the three
6:16 pods that we have set up is shown up
6:20 over here. Now these pods represents
6:23 each and every pod that is associated
6:24 with the aerospec cluster. The reason
6:26 being we had asked for three pods to be
6:28 set up over here. Now what we are trying
6:30 to do here is to access it and show you
6:33 that these environments can be accessed
6:35 using uh the ACDM command which acts
6:38 like a client. If you're able to client
6:40 use this client the same thing that uh
6:42 goes with the any other client whether
6:45 it's Java, Python, name it we have we
6:46 can connect to this particular
6:48 environment using these commands. So
6:52 this asadm gives you the value. The
6:53 admin user that we have set up as part
6:55 of your rolebased access as this control
6:59 gives me access to it and whatever we
7:03 have set up as a YAML file gets set up
7:05 as an environment. AO cluster is my
7:07 cluster name.
7:09 E6.1.0.1 is my cluster value and the
7:11 cluster size being three right now. So
7:13 this is just a first part of the story.
7:15 Now much more complex things. For
7:16 example, you want to do something like
7:19 an upgrade of an existing environment
7:21 and you don't want to spend most of the
7:22 time. So what we have done right now is
7:24 we are trying to upgrade this
7:28 environment from 6.1 to 7.1 as well as
7:30 to increase the size of the uh the
7:32 number of pods to four and we just go
7:34 ahead to apply this particular value and
7:36 that is what is being showcased as part
7:40 of this demo and how seamlessly uh this
7:42 gets migrated to the next upgraded to
7:44 the next version in a sequential fashion
7:46 or a rolling fashion is what you'll see
7:48 on a window on your right hand side.
7:50 Again, we have applied the cubectl apply
7:52 command with the new YAML file that we
7:55 had right now. And that takes you to the
7:57 next window. So, what I've done is I've
7:58 just copied the configuration to
7:59 increase the value. And that's what
8:01 you're seeing over there. And here's
8:03 where we are applying the same YAML file
8:06 that we had initially triggered. And if
8:08 you see there, it's showing us
8:10 configured. And that now we take you to
8:12 the right hand window wherein we are
8:14 doing something like an info network
8:16 which tells you what sort of build was
8:18 installed over there. Now if you can
8:20 concentrate on this particular column
8:23 over here called the build migration the
8:25 build E610 that's what we started off
8:27 but what we have put up right now is
8:29 7.1. So what's going to happen over here
8:31 now one by one you will see that each of
8:33 this node will be brought down we
8:36 upgrade it and we are glad to announce
8:38 that we support something like mixed
8:40 clusters wherein we work with both 6.x
8:42 and 7.ext at the same time. So without
8:44 any sort of downtime in a rolling
8:47 fashion we are able to migrate from a
8:50 older version to a newer version of uh
8:52 aerospike and that's what you're seeing.
8:54 The first one is already through. So it
8:56 waits for the migration everything that
8:58 you are expecting from a end user is
9:00 automated as a best practice within the
9:02 AKO. Now if the first one is already
9:04 done then the second one will follow
9:06 suit and that's what you see by running
9:10 the info network again and this takes
9:11 couple of seconds or or the time it
9:13 takes would be the time for the
9:15 migrations to complete. So you'll see
9:17 the second one going through right now for
9:18 for
9:20 migration. It's up. We are left with a
9:22 third one. So for for the time being we
9:24 have just two of them already migrated.
9:26 The third one will follow suit shortly
9:28 and the next step would be to increase
9:29 the number of parts. Initially we had
9:33 given it as three and now it'll get uh
9:37 migrated. It'll become four parts very
9:39 shortly. See this is the end result that
9:42 you wanted. So seamlessly uh without any
9:44 sort of manual intervention using the YL
9:46 inputs that a single place from where
9:49 you can run the entire upgrade process
9:51 and finally the fourth node as you can
9:54 see is being spun up. So this sort of
9:57 automation is what you get using AKO and
9:59 every other best practices that we have
10:01 has already been built into this
10:04 particular tool. So this is about the
10:07 demo. So first question that comes to
10:09 your mind is there are a lot of moving
10:10 parts right has this ever been tested by
10:13 yourself or by any other client of that
10:15 you know of. So we are glad to announce
10:17 that we have one of our part key
10:19 partners which is Flipkart who has been
10:22 through this journey and uh I think is
10:23 there folks from Flipkart who can raise
10:25 their hands
10:28 please. Yeah, great. These are the
10:30 wonderful champions like FI membered
10:31 superheroes who are managing somewhere
10:34 close to 300 odd clusters with zero
10:37 downtime and the latest sales day they
10:39 were able to uh manage somewhere like 35
10:42 billion transactions uh over over this
10:44 particular 300 cluster with zero
10:46 downtime and that's the beauty of how
10:48 simple is it and how effectively the
10:50 team over here has been able to manage
10:52 it as well as we have also have critio
10:54 as well which is another customer which
10:56 has almost the same range we have many
10:58 more customers who have done this right
11:02 now and aerospike has your back. So
11:04 ultimately we have our engineering team
11:07 which is all ears to see what how best
11:09 can we evolve this product further. So
11:12 we have got excellent interactions with
11:14 the individuals who are already using
11:16 AKO and we hear them to make sure that
11:18 the subsequent releases has those
11:27 So from a maturity model standpoint
11:29 where does operator stand right from a
11:32 maturity standpoint we are at level
11:34 three wherein we are working on creating
11:36 something like a backup failure and
11:38 recovery built into this particular AKO
11:40 and by end of this September you should
11:42 be able to see that these features will
11:44 be there in AKO number one as well as we
11:47 are moving our sites into the next one
11:49 wherein we want to do deep insights on
11:52 how do we metricize how do we read these
11:54 metrics create alerts, meaningful alerts
11:56 and make sure that we process these logs
11:58 which in turn will lead to something
12:00 like a autopilot mode wherein you want
12:02 to scale up much more effectively,
12:04 automate the entire flow, all those
12:07 things will be part of level
12:09 five. So that brings me to the next
12:11 slide which is the benefits of ATO.
12:13 After seeing this right now all these
12:15 the wonderful things that's moving
12:16 around you can make sure that you can
12:18 have a automated deployment scaling and
12:20 management much easier. That's number
12:22 one. managed by the centralized team
12:25 just like Ash for show showcased by our
12:27 colleagues over here as well as it leads
12:30 to higher engineering model because 300
12:32 400 odd clusters being managed by five
12:35 individuals is not a a joke right that
12:37 these guys have been able to do that and
12:40 it leads to a higher engineer morale and
12:41 adherance to standards being managed by
12:43 a centralized team brings in something
12:45 like an adherance because they know what
12:46 are the best practices that has to be
12:48 involved so they make sure that that is
12:51 adhered to and finally Every other thing
12:53 has got a positive business impact right
12:55 like effort that they save over here can
12:57 be channeled towards in innovative
12:59 efforts which in turn leads to something
13:02 like a dollar value SLA is being met
13:04 like what's being showcased and zero
13:06 downtime is what you will definitely get
13:08 from an aerospike which which is taken
13:10 over here so we are glad we have our
13:12 booth over here which with any other
13:13 questions that you have we'll be glad to
13:15 take it up and as I said this demo is
13:18 just a teaser there are much more
13:20 complex things that can be managed We
13:23 have a host of uh GitHub repository CRDs
13:25 which gives you entire combinations that
13:28 you can try out like XDRs making sure
13:29 that you want to change change some
13:31 parameters those are all available.
13:32 That's about it from my end. Thanks for
13:35 listening and have a wonderful day. [Music]
