Cloud Native Computing Foundation Kubernetes Forum Seoul 2019, 10 Dec 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Improving Monitoring Systems Interoperability with Prometheus & OpenMetrics- Chan Shik Lim, NexCloud

Description

Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects

Improving Monitoring Systems Interoperability with Prometheus & OpenMetrics - Chan Shik Lim, NexCloud

Prometheus is the top-listed monitoring technology in use for container clusters whilst enterprises are taking their legacy IT infrastructure the cloud native environments such as Kubernetes. Prometheus collects and stores the metrics that are preset in exposition formats that it makes the container cluster monitoring feasible by exporting the metrics data in exposition format regardless of OS and programming languages.

The OpenMetrics Project is in progress in an effort to further extend the Prometheus exposition format by determining a standard for exposing metrics data, and is aimed at allowing heterogeneous monitoring systems to share data effortlessly as a means to improving interoperability. In this presentation, ChanShik will cover the methodology of improving Prometheus-based monitoring systems by taking advantage of OpenMetrics-based standard metrics protocol.

A

A I've been signaled to start. Hello,, now I'm going to.

A

Show you how to do it, while talking about the intermediate medium rope matrix about how to use different monitoring systems together.

A

Interoperability and accusations using arc matrices, 5, Prometheus, A, more detailed explanation of the arc matrix is different. I'm.

A

Always looking for a developer. Let's go ahead and show you a demo first.

A

Yes, in the end,, my mother has everything: I want to say.

A

First of all,, the environment is like a screen. The vm.

A

Is m after running the program to imitate the actual iot device. For, this vm, a program that operates like an ik application and an agent that collects matrix information subtracted from the application are two agents. And. These two The agent collects the mat rig of the iot application and transmits all of them to the server running on the cloud platform.

A

So I put it in the open matrix of the application, that is,, because the current from Prometheus is so different, I.

A

Added a setting to subtract the metric in the form of an off-matrix later., I 'll show you the appearance, Other.

A

4 Currently,: this is the Mat League collected by the iot application in the vm. The.

A

A

Application is the Mat Rip information given by the running program like the iot application. In.

A

This way,, the two applications a are all matrix collection,. So the.

A

Current content is e. The.

A

A

Is set up like this to collect a lot of mat rings that meet the specific conditions without collecting all the netflix 2 exposure. Tricks. Also,. If you use the Protheus client libraries,, they basically give out a lot of metrics. They, also give out information about the process. If you put in each language and do it in a language,. Flowers are c-related things.

A

Then, all the information collected in this way is sent to the server. Since. It is currently on the above notebook,. It is absolutely difficult in 5 Prometheus on the cloud platform.

A

Therefore,, it will not be pushed or released, and it cannot be done. Illegally. Yes,. If this is done,, the server will re-expose the mat, rig. The metric information exposed in this way is used by Prometheus to travel,, and the data is also re-exposed through the open matrix integration of the agent.

A

A

Show you again,. This server has two exposure: tricks, Device, One and Device to Application. For example,. If you look at the label,, you can see the metric for the current device. Tool.

A

Yes, the server collects and re-exposes the metrics that were sent together from the device one-two.

A

Since every trigger is not different,. These are the methods that are basically used. A lot. I installed two open matrix inches of data on this and added the settings to take the metrics on the server. It's.

A

Showing matte lip information. 4 It's, not that the open matrix is very special or anything like that,.

A

It's the current cn chef sandbox project to re-evaluate the standards to be used between monitoring systems,. So it's a project that was born for that purpose. The.

A

Hood for the matrix was installed in the program,, and the matrix was exposed in that way, and by using it,, Prometheus and data. Besides that,, there is probably still a search.

A

So, they already provide a function that can be output through their monitoring, dashboard.

A

4 Go back to the slide,, and perhaps the description of the Mat Purification Pu matrix may be a little insufficient,, so I will explain it a little more. I 'll start with the explanation. First. Sales special is such a wiki project that is currently called Chef Jin's graduation, project. If. You.

A

Focused on presenting and showing it, you.

A

Have provided a data structure in the form of multi-man journal data,, which uses mat, ring government and labels to show data in an expanded.

A

Form. Then,, Phil and Press Manager provide such a function that can be used to query data, unroot, and send an alarm. You.

A

Should be familiar. It's, not that different here. Just a little bit more information about the metric. Prometheus battles are similar to storing the key value stores for gold, dismantling. I.

A

Will now specify the tap value in the form of a plot, 64 form. In, the metric integration juice,? Now this is a bit of a format for exposing metrics,, but the metric, name,, label, and time stamp are basically put in.. There is a histogram summary of the counter gauge. Yes,. You must have seen charcoal,, which explains the taste in detail.

A

Does the counter keep increasing? Beige is about changing values, or things like this. It is recommended to indicate the domain. Name. At the end,. It is called the Uni-Friendly Unit,, which is called The, Byte, Second, Total or Info. It.

A

Is a specially treated infrastructure that is used to indicate something like an application version or agent version rather than a metric government. Yes,? The label is actually a feature that allows the metric information of Prometheus to be used very flexibly.

A

Using the label,, even if it is the same matte lip information,, it provides different metric information, conveniently. For example,. There is a battle called hp1 quest, total By, making the path that enters s different. When this is hp,. This total work can be used later to retrieve metric information when querying using this, or to filter, using labels when blocking, or to leave values using this or to obtain averages.. One thing to note when using labels is that, as the number of labels increases, data space continues to be occupied as the number of types increases,.

A

A

Not recommended to use IDs, e-mail, addresses,, money, and other data that can become very large as labels. The value is called a sample, and it's a lot. 64 value., Emperor, timestamp or Prometheus uses millisecond timestamps. Or,. It's one of the things that differs from the arc. Matrix. This is how we talk about mat 3 with 4. Let's talk about the Fu matrix. I'll. Do it First of all, before Prometheus is now open to the public by itself and used,. There.

A

Was one effective way to use snmp between monitoring systems on the monitoring? Side. Those who deal with the network, probably have heard of it at least once, and probably routers or routers. Those, who have worked or worked on. Such things must have seen a lot of snmp. It.

A

Was almost the only standard, 4 type. It was impossible to exchange data between monitoring, systems. It.

A

Wasn't easy,, so this is what monitoring systems had before Prometheus. After Robe,, exposing metrics in text form is very widely used. And.

A

The program has its own way to expose metrics with Prometheus. If not,. These applications, called exporters, extract the metrics of the program and expose them. For example, in the case of My Scale or a database like this,. Now you can get the metrics. If Perl sends you something.

A

A

Subtracts the metric in the form of recognizable text. There are hundreds of x porters like this.

A

Monitoring systems are now moving towards using vape hair as it is, so.

A

It is widely used. It's a little bit like that,, so off-matrix started the project and a lot of developers participated in a lot of companies, developing monitoring, systems,. So I thought about how to do this in the form of fusion or something.

A

So,. It is not dependent on the vendor that wrote it,, but it is.

A

Registered as a cfs box project to achieve this, and in conclusion,, it is currently in progress.. It is still in progress and the following: year, for example, ssh writes h2 80 times by itself. 22 times, and the port is also designated, so that it is working hard to establish a position like a real standard., 4, Prometheus, laughs and 70. There are some people who are sleeping a little.

A

Uh basically,. The unit is active and displays what the matrix will display. Next. At the.

A

End, I made it explicit that f was displayed. So,. If you meet this application on the side of the taker,, it is all over. Why. Is this basically Prometheus?.

A

I'm trying to support something like this,, so it was created by adding people who requested it. Earlier. The timestamp changes to second. It's.

A

Been discussed to support both the pour and push methods,, so that means the basic cooling method,, but Mr. Hu is also being discussed.. Because Monitor 4 systems have a lot of push, and the content of the photocall buffa continues to be discussed. Yes,. It is a binary protocol with text.

A

Let me show you a little more. For example, in the past,, when using a counter,, an explicit underbar total was suggested. If, you make a claim,, you.

A

Are making a total and k-kid together, or off-matrix in this form is now being proposed. It.

A

Is a sin to make it possible to leave 10 or 1 link information to make it easy to track.?

A

This is probably the best among the things presented in the current Woof Matrix,, which will cause problems with Prometheus, which collects existing Mat Rip information. Now,. We are talking about 3:00, timestamp. In, this way, after the Matrix has taken off the name of Prometheus,. The project called Matrix is progressing,, and it is now in progress to establish itself as a standard among monitoring, systems. To. Take a.

A

Little bit of data vita dock as an example,, the data is already included in the matrix after.

A

Two inch migration.

A

After that,, save it on the server and.

A

Provide it as it is in the main monitoring system that you are using,, such as Prometheus,, Bella, t, Seojin, Star,, or something train,, etc. in the form of a parts matrix,, so that you can change the monitoring system or do something. When using the system,. In the case of other things,, we are proceeding with the play in a form that can be used together without removing the existing one or doing this. It's.

A

Out there,, but the translation is out a while ago. The author of Asai's work is now Brian from Prometheus's country, and he's contributing a lot to Open Matrix. I.

A

Used to use Prometheus quite a lot. I'm very curious about foreign countries, because basically, there are not that many monitoring systems available for free,. So maybe these days, Edith 9 is good and used a lot. It's.

A

A

Possible to get out of it. Yes, rather than below,, it was announced at the time from Promphone. Now,. The next thing that came out is run.

A

Stocking. We did it, and we went through.

A

Trial and error. It.

A

Is difficult to collect in the case of disconnection. It is difficult to collect such things by using things like Tanhus or x. I.

A

Saved the information,, but since the open, 4 trick itself is not for one monitoring system,, but for multiple monitoring. Systems, I changed the default timestamp to seconds and saved it in milliseconds, using a unit or something like this, for example. If. You want to,, then use the Unity Lek skin to change the explicit designation to this form. 1 Because. The existing monitoring system is the smallest in the second, and in the case of Rafale, it is difficult to repair everything because of the deep matrix. There is.

A

A

Pair Once, you use the Pro Mat in the case of No. 2 Tis No., 2, I will probably use the Ham chart. A few times. It is.

A

Recommended to add a label using , so that the field is automatically collected. If. You have any.

A

More questions, in a way,, it is still a cent box,. So if you actually go to save 3,, there is no code. A matrix parser or something like this is still being developed,, and if things go, well, I'm really looking forward to taking a place next to snmp as a standard for the monitoring system.

A