Cloud Native Computing Foundation KubeCon + CloudNativeCon North America 2020 - Virtual, 4 Dec 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Observing Cloud Native Observables with the New SIG Observability - Bartlomiej Płotka, Red Hat

Description

Don’t miss out! Join us at our upcoming event: KubeCon + CloudNativeCon Europe 2021 Virtual from May 4–7, 2021. Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.

Observing Cloud Native Observables with the New SIG Observability - Bartlomiej Płotka, Red Hat & Richard Hartmann, Grafana Labs

This year we founded a brand new SIG to take cloud-native observability forward: Welcome SIG Observability! Consisting of open source project maintainers, standards authors, end-users, and more, we cover a wide range of experience and invite you to join us as well! In this talk, the audience will learn what our goals and objectives are and how anyone can get involved and contribute. We will give you a short history of what we did up to now, what we’re doing right now, and where we want to go in the future.

https://sched.co/ekFx

A

Hi and welcome to observing cloud native observables with the music observability, which already tells you that observability is kind of a buzzword, but there is some value to it. I'm richard I'm bartek and we are going to tell you what this is all about. Sig is a special interest group and observability at a very fundamental level, is about making complex systems groggable for humans and that's basically what we're working on. So, if we look at our charter, it's very much around fostering reviewing and growing the ecosystem off and around observability.

A

Identifying and reporting gaps within the cncs project, portfolio find establish, write down patterns and best current practices, and the current is important because things change over time, educate and inform users about all of these without any bias, which obviously means no vendor lock in a truly vendor-neutral venue for all thought, validation and such and as a little bit of a test balloon.

A

um We are doing more in in due diligence reports than other six, where we actually walk through the whole due diligence process to take some load off of toc and one thing which you really really really try and do is have a bias for action, because usually um any any new endeavor um is always nice and cool and shiny, and everyone wants to join and we're trying very, very hard to to actually push for actual bias for actual action and virtual outcomes.

A

B

A

Activities we obviously uh chartered ourselves and self-created with the blessing of the toc, and we did already two incubation reviews and due diligence reports for cortex and thanos, both of which passed and are now in an incubating stage.

A

We started gathering requirements about online learning, online, analytics platforms and analytics, and we already explored some overlap with ai and ml of course, especially in metrics land. This is like perfect signal for for machines to understand and grog, and also integrate with other projects for some knowledge transfer.

B

Amazing and as you might be already aware, our observability landscape in the cncf space is already quite quite rich in we have kind of different levels of um of project stage within the cncf and uh in graduated we have you know, fluency, which is related to logging jager, which is a tracing system prometheus, which is the first system that that kind of get into the cncf from the observability space around metric monitoring.

B

And then we have, you know incubated stage where we have cortex and tunnels, which are you know, metric um kind of metric systems, scaling premiums and those are kind of bolded here, because those are where kind of moved where well, thanks to the seek observability kind of movement, and we are looking to help- you know other projects to graduate as well and join our space.

B

We have open tracing, which uh now kind of merged into open telemetry, but still is incubated project around tracing and then in sandbox we have, you, know, cow smash and litmus for for ready to do cows, engineering and uh yeah. Please join our. You know, meetings to learn more about them. We will kind of share knowledge as well.

B

uh We have open metrics, which is um essentially metric format, um kind of exposition, uh well-established format and protocol around around all applications and open telemetry at the end, which which has amazing progress, as you could see, if you're following the twitter or or actions- um and you know kind of collaborating with many many vendors and and allowing them to kind of um integrating well within single, simple kind of application, you can you can put on your infrastructure, it's kind of amazing.

B

um So you know the observability kind of uh space here is is very um rich already you can. You can do a lot just using those projects, but you know that's not everything right there. We are definitely not done yet with this landscape and there might be some observability areas that you know are missing uh from this landscape, or maybe we can integrate better with right.

A

So if you look at the integrations or at the overlap, there is obvious overlap between the three golden signals: metrics logs and traces, in exactly this order, uh basically pareto principle, which that's just the ordering of the most bang for your buck versus where's.

A

The effort you need to put in so it usually makes sense to start with metrics then go to logging, then go to tracing, that's just kind of the natural evolution and, if I do say so, myself, one thing helped really connect the dots a bit um where openmetrics took the label based metrics exposition format of prometheus and already back in 2015.

A

The the intention was to also have logs and events with the same label based access mechanisms which, with profana loki we have, which is super nice and, as was announced, uh we're recording this on a tuesday. So yesterday of my time, I don't know by heart when, when this is actually showing um tracing based on exemplars, what that means is basically you only have an id for your traces and nothing else. So you come into your traces from metrics and logs, which are relevant to you course.

A

They have like high latency, or there was an error, or what have you and those obviously carry label sets with rich data which enables you to actually go to what is interesting to you, and then you just jump into the trace with the help of an idea, and you don't need to have that usual needle and haystack problem of of tracing and personally, I think that examples are the future for traces like absolutely 100, and this is why even back in 2016 openmetrics already had support four traces which permeated through the prometheus cortex thanos and also loki ecosystem and with grafana temple.

A

We now have something which actually stores traces based on on examples, but there's more, we have continuous profiling, where you have an obvious overlap between metrics and tracing, which kind of goes in this direction. But where does it sit precisely? Should we have a new project which which can actually store those, or can we make either of the other fields of or groups of solutions? Work for here then there's another thing which I think is too often overlooked. Crashdoms.

A

You can argue that there is somewhere sitting in between tracing and logging, but having informed state about hey something actually aborted abnormally, and this is what what its dying gas was. I think will be useful in the future to actually figure out why something specifically hit the wall. But this is just us thinking about those things. There are certainly other signals and we would love to hear about your thoughts around what else might be missing here, because that's that's literally what aesthetic is about.

A

It is a special interest group and we highly highly highly recommend and invite anyone to come to us and actually talk about what you want to see in this space.

B

Yes- and there are actually you know much more- we are doing right now, which is on our plate. um So, let's, let's quickly, go through that um we started. You know our proper meetings, let's say a couple months ago, really so so pretty uh pretty early. um However, we already started kind of a few directions uh here and there- and it's really about you- know, building this community around observability.

B

To make sure you know everyone um can contribute in their of um expertise. So you know by allocating those kind of small working groups- and let's say some groups that are focusing on one area: we can. um We can just scale and do much more better together, so um examples of things we are doing right now. We are definitely looking forward to just start. You know, seek observability strong and to learn more about projects and each other. That's why we are planning to have some introduction webinars.

B

We already have some introduction, um thanks to steve from about open telemetry. We are planning to, um I think, host litmus project as the first first one um which will introduce themselves. So uh we can learn about. You know their problems each problem, each project's problems- maybe we can have uh you know kind of um collaboration there in terms of um you, know, integrations and kind of user experience of of of of really any concept of maintaining open source product, even that is related to observability.

B

So that's our plan here um and uh other aspect is, you know this lacking of well integration between a good integration between, let's say infra, infrastructure engineers and big data, analytic, analytic engineers. um There is, you know not much. You know conversion formats and api integration towards this connection. So, um as the first thing, we started to yeah look on how to get metrics from uh how to get metrics from from the projects we support like from to use, thanos cortex and get all or like open, telemetry and get into um target file apache arrow.

B

You know many many projects from apache foundation, for example, how to integrate that so yeah we started some poc, some cool coding, but also some user requirements. So, depending on your background, you can help in many many areas uh next, one which is kind of starting right now. We recently had this observability, rather, which was amazing, because we, you know, suddenly have more data to understand what is the interest uh regarding the projects and the problem space.

B

However, obviously you cannot, you know easily ask people, um you might have you know some biased information based on your connections and that that's normal, so we are kind of starting some effort into making. You know maybe uh improving this, this kind of um surveys and making them you know easier and and actually make companies motivated to actually take part in those surveys to actually have meaningful data that everyone is um is trusting right.

B

So there is some work on that side um and last, but not the least, is introduction to observability right like there are lots of questions. There are people coming to our meeting and saying: okay, I don't know where to start. I just you know want to improve observability in my company. How do I start, and there is no good like you know, starting um starting docs, or I don't know some index that will root you to the particular projects or materials that explain you. um You know certain certain problems like for example.

B

How do I gather metrics? When do I switch to locks? uh Do I need tracing? What is this router tracing? What is profiling and all of those are maybe for some people, it's obvious, but sometimes some direction. Initial direction is very needed, so we are looking um forward into into creating some ways of having some, maybe a good introduction page that will root to at least um you know, cncf landscape projects.

B

um I think it's a good opportunity to also collaborate with other sikhs and to to learn how they share knowledge uh in and and have some method in doing that. So, if you want to help in this direction, uh maybe help us writing some some better documentation.

B

uh Maybe you know uh actually write better documentation on each of the projects and and kind of root and link them together in one place. That would be super amazing and we are looking for your ideas and feedback and last but not least, yeah. We want to um help projects to graduate due to further stages. We want to make sure they are well supported, well maintained, so um we are looking into you know, uh proposals of of getting projects higher, one particular ideas or like proposals we will be working right now is open.

B

Metrics uh we are moving, I mean openmetric is moving quite fast and we are considering you know. um Yeah maybe have some review in terms of incubation graduation. So um all sorts of kind of actions like that is in our scope.

B

uh Next, I'm bartek, I'm kind of you know a tech lead here and we have richie here who is sick, a chair. We have also matt young, who is another sister, and this is kind of a starting point, but we have also, I mean huge amount of amazing people who contributed already to our sick observability. They were on our meetings. They help us with the review and preparation of the charter dogs, and they are part of you know all those smaller working in groups and direction.

B

We are, we are creating so um yeah, and so many of those and and it's super nice that you know we can kind of build the community with those amazing people and and um anytime. You need help. They are super happy to just you know, use their time in order to uh and reach this observability. You know open source. um It's super amazing and thank you, everyone and yeah. You will meet all those people, although all of them are friendly and amazing. So uh please join us just to meet people. Really!

B

That's that's one first um argument you can have and yeah um so action action item for you how you can help how you can join us if you want well. First of all, we have b weekly meetings. um You have a google doc here, we'll share slides, so you can uh click on it and we focus on those meetings to be.

B

You know mostly introductionary, maybe some uh presentations, maybe some status update, but we've meant for those meeting uh um well for the actual work to be to be done outside the meetings we are to to to large group already to actually you know, do everything in one hour. So um please be prepared for that, but obviously please put stuff on agenda and and add you know any topic that you that you want there. We have our repository go there. There are like basic informations there um put an issue on the repository.

B

If you, if you want to uh have some asynchronous method of of doing some requests for us, uh we have seen slack, we have main link list and yeah you can help with all of those stuff. Maybe you have some feedback? Maybe you want to have. uh Maybe you need some help as a project of on observability project or maybe observability project, maybe outside of cncf? Even you don't know how to start how to integrate. Maybe you have ideas, we want them.

B

Maybe you have just free time and you don't know what to do with that. We will definitely have some ideas how to allocate this. So please join next. So that's it. um Thank you for for listening. um Now is really time for any live questions you have to ask. Our talk was a bit shorter just for this purpose. uh If you want to take this opportunity and ask questions related to seek observability and just learn more, we are happy today, question questions.

A

Yep, looking forward to the questions, thank you very much.