Grafana UX Squad, 30 Jun 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Grafana UX Feedback Call 2021-06-30 – Hackathon Edition

Description

It's Grafana's first hackathon! We have united to invent, build and solve challenging technical problems related to our products, community and business. Some of our hackathon ideas were discussed at this week's UX feedback session, and these are: Configurable kiosk mode, Service mesh view, and Jupyter notebook.

A

So who here is already working on the hackathon project? I think marcus is right, who else project teddy awesome, but do you want feedback on your hackathon projects? Yes, oh really. Okay, that's good!.

A

B

um Me and patrick were in the in the same hackathon team, so I think he might have prepared some some cool things.

A

Okay, all right we're, I guess we're two minutes in. I think we can just get started.

A

um So I guess I can take some more. I could take notes, um so it's mentioned here. Configurable kiosk mode. Is that the same that you're working on marcus, okay, cool who's going to present?

A

I guess I am awesome all right. Please take it.

A

C

No, no, I think right.

C

I always find it complicated with zoom and sharing here.

C

Yeah, so the idea is sort of an improved kiosk mode and also making it configurable.

C

So at the moment, if I would go into kiosk mode now, I would still have these buttons here and these buttons here, but since cross mode is more used for having on tvs on the wall, it's not something that you really interact with. You should do something more a lot more visual information. Just the initial idea here is to basically when I click kiosk, what mode will be left with is the title of the dashboard and then the time range and how often it refreshes.

C

And what we have been talking about is that if I would move the mouse and these things stop, the menu would come down while I'm moving the mouse. So then I can sort of configure things if I would need to and then, when I start moving the mouse they should vanish.

C

So that is the first thing I would like some feedback on if that makes sense or yeah, and the second thing is then configuring it. So then you would go into dashboard settings and we would have just mode options from here. You would actually show the options for the standard tv on custom, so these checkboxes would be disabled, so you could not change them.

C

It's just to inform you uh what is being shown in this mode and then you would have tv mode which basically has hidden everything except the panels, and then you will have a custom mode where you can sort of pick the things you want to show on your dashboard um yeah. So that is basically it. The issue is here: these things can grow. We are not sure how much stuff people will be able to configure here, but this is sort of the initial uh exploration here.

C

So this is quite rough um mockup, just sort of showing the functionality- and I don't I mean the animations here- are ugly, so we don't need feedback on those yeah any.

A

Comments who wants to do it first.

D

um I have, I don't, have any feedback, but um I did talk to a lot of users who use grafana for more industrial use. Cases like they'll have grafana up on the screen in a manufacturing floor, and so this would be really helpful for them. So maybe in the presentation you can like list some of those use cases. I can share some research that I have from that.

C

It will read super.

A

Yeah anyone else.

A

uh I I've got feedback, um essentially, if you show the tick boxes, if you show the check boxes and they are disabled, I feel like could be a source of frustration, so I think one trick you could do is as soon as someone clicks on one.

A

Then you immediately switch to custom and it'll derive from whatever was pre-selected before.

C

Yeah or could they the options? Basically, the settings be presented in a different way, just like.

B

One thing that I thought of that: I uh you just now when you did this demo, if you go back to the settings page, maybe it would be nice to have a description on each item as well or or maybe they are self-explanatory. I don't know.

C

um I do not think there's a explanatory to that it would be dependent if you sort of I mean they are sort of. When you see what is yeah, so we probably have like a proper description, we're describing what kiosk mode actually is and what the different types are. So that could be.

A

One, if you want to go all out, build the best key kiosk option editor ever you would show a preview right, yeah.

E

My question is uh like, if you choose the dashboard header, does that include like the time range picker and like the refresh thing like oh.

C

Yeah I should have yeah. I probably should rename that to dashboard title. That is a title. Yeah.

E

C

uh But uh yes, so the interaction of showing the menu, the toolbar when you move the mouse does it does not feel like it's a good thing or is it something anyone has anything against or cpr issues with.

E

So I I thought about it from like an accessibility perspective there. So if you, if you would use the keyboard instead,.

E

Maybe then like, if you have an element focused, uh then it could also be visible.

C

Yeah definitely underneath yeah. It should sort of be a similar interaction with the keyboard as well. Of course. So it's.

D

C

Like because I think the use case is that you set it up and when you put it into kiosk mode, you don't really mess with it. That much.

E

More yeah, no, no exactly.

C

Yeah two more minutes: if anyone has any.

A

Comments all right uh well, we'll have to do the full two minutes.

C

uh But if, if nothing more, I can stop sharing and we can go on to the next topic.

A

um Should we go straight to olof, where is it off.

E

Oh at the bottom there I'm here yeah.

F

Okay, I don't know if there's anyone else that wants feedback on their hackathon projects. I guess they can go first.

A

Anyone have any headphones.

E

A

Some hackathon.

E

Products but like it's.

D

Yeah, I can share something it's I feel like the the cool parts of it. I haven't gotten to mock up, yet I've mocked up just the very basic bare bones aspect of it, um but maybe I could get some feedback on the idea, so um I guess, as a as context, we're building a basically like a service mesh view for kubernetes services and apps.

D

What we wanted to do was sort of break down the cognitive load that it requires to sort of visualize all of your services and apps and the hierarchies and dependencies and show them in a visual way, as sort of like a default, automated view that you get as a part of grafana if you have kubernetes um and we want to be able to show sort of like a status for each of those services that at first might be just like up or down, but in the future we're thinking it could be customizable like um being able to write out a query um or like a composite score that would be comprised of multiple metrics and thresholds and weights associated with those.

D

So um I started mocking up views of what this could look like and I haven't gotten to the cool views. Yet I've gotten to the boring views, so any feedback on the idea or the visualization could be helpful um here we go. Can everyone see.

D

All right, um so we were envisioning that this wouldn't necessarily be a dashboard. It could almost be like an application.

D

um Maybe this actually says mesh, because that's more of like the industry standard, um but you'd have sort of like this view, where you could switch between a list view a card view and a visualization view to view all of your services, and here we've just we've made the decision to list out services first instead of applications, because then you can expand this row to be able to see the pods that are associated with that service and also the status of each pod, um and we didn't want to have like a multi-level drop down where you have to expand the application and then expand their service.

D

So um this is sort of how we've organized it and you can filter by labels that might be helpful to filter we have like owner and team and service. Maybe these are labels that you've applied and also the ability to to filter by application, and we stole gems status ring so right now. This is showing basically like. If cart api has four pods.

D

It's showing that three of the pods are up and one is down, um and then, when you expand, you can be able to see which pod is down um and then we're also thinking it could be helpful to have sort of like a service drill down view. Maybe you click into it and you see more information is like a dashboard view where you can see error logs associated with the pod. That's down.

D

um I've been mocking up other options for this, like this would be using our cards component instead of the table component, um but I don't really think that this adds any value and in fact the table is more nimble because you can have sort of like the expand collapse rose, and then we started playing around with um what the visualization could look like.

D

So this would simply be um sort of like a card view of um services, but what we'd like to show is actually the traces and dependencies between them and so um I've been exploring.

D

This was another option I had like. This would be like a short time series of if the pod was up or down, but I've been looking at. Maybe these panel types and then also the node graph, but would love any feedback and also something that um is good to note is that for this view and being able to show the dependencies, um we okay, sorry, two more things.

D

So we like having it as more of like an app because it's it's like this central location, where you can see this mental map of all of your services and then we also have like a configuration tab and a how-to tab that I haven't mocked up yet, um but in here in configuration you could maybe set a custom composite score and custom thresholds and then in the how-to we would actually have sort of like in-app uh documentation about how to instrument your services um using kubernetes and whatever tooling, that you have to be able to have a cool view like this.

D

So, for example, adding labels is really helpful. um So we would have sort of like a step by step here, that's right in the app, so you don't have to go and like leave the tab and start to figure out. How do you you have to instrument things to make sure you this view? Would work for you um and so for that reason we're struggling with like how would we pull in almost like a panel type into just a view? That's not a dashboard!

D

So sorry that was a lot but any feedback on the idea. The visualization implementation, like.

D

B

Really cool project- I really like this and one thing so.

B

When I previously been working with this kind of systems and usually in the like in a visualization view, uh what I usually is aiming for to to get like as a overview is maybe the status of the like my system, so, for example, if if my my system uh is built in this case, you have a couple of front-end services, you have a payment services, you have a card service and you basically just want to get an overview of.

B

What's the status is everything functional and then, if something is broken, usually you want to see, as you said, like the dependencies between them, so you easily can see like okay. So we have, uh we have the the card. Api is red, the card db is red and the payment api is red and all of those two are all of those three are depending on a like another service, but that one is green or that one is yellow or something.

B

Then you can probably figure out okay, so so the issue is probably in the the yellow service to start investigate there. So you, like you, get you get like this overview. Where you easily. Can see where the problem probably is and, and that that that's some like so so the dependency visualization, as you said, would be pretty nice like comb, combining that with some kind of color for service, as you have here.

B

And okay, I I really like the the status indication you had where you have like this. You split up the the pie, basically in different parts, to indicate like how many of the pods is green and red and yellow and really nice.

D

Yeah for the dependencies we were looking at a tool called kiali, and they do it really well, they sort of have this like visualization of different services, and this would be sort of like the application, and then you can show how they talk to one another. So is this more like what you were envisioning and then they'd have actually like a this. The color would indicate status for each.

B

Yep, that would be like spot-on.

B

It's it's kind of tricky to build. I think, but it's uh if you have it and make it work, it's it's really. Nice.

B

Because I I think that gives the user like a really quick overview of how your system as as whole, is functioning and then you can pretty easily figure out. Okay, some something's bad is happening in this part of the system. So let's just jump in and investigate stone's part a little bit more and the thing with the dependencies. Then you can probably figure out like where in this dependency chain is the issue.

D

G

Said really nice thanks for.

D

G

um I have a couple of comments. um You showed the how to tab where um you're thinking of adding some uh documentation uh shouldn't that help, or should that be just like a question mark, because I haven't seen anything um similar in in our other. um You know other parts of the ui so wondering about that. I wasn't like very clear to me um and I'm sure you know you'll be thinking about this going forward and the other thing is like the labels.

G

uh You showed like two different mock-ups one had um the same color orange for all the labels and then the other one had different colors. I think the the one with the different colors probably makes more sense. It's like visual, exactly like what kind of labels are added to each of the um clusters that that you are showing um yeah cool. Thank.

H

D

A

uh Yeah, I would say it's pretty ambitious project. um I think if you get just any of the views done, I think it would already be a big success.

A

um Yeah, um I think maybe one really interesting thing would be to also hook this up to one of our kubernetes clusters right. So if you could do that, maybe as early as possible to get a sense of what kind of oh like how big these like normal sets of entities are like how many applications, how many namespaces, how many services are sort of normal or like? I guess I don't know- I don't know if you're normal, but I think the dog fooding aspect is really cool. Still yeah.

D

Yeah yeah, that's good advice, we're using um sort of like a. I don't know what to call it, but like a package of like a fake um application and services called sock shop and it's all fake data. So obviously it would look really nice here, um but that's a good idea um and then, as far as implementation goes sort of what we were planning was like um this view.

D

I think um kevin and andrew are building, and then what we were going to do was like show mock-ups of where we would take this idea, but maybe just build this to begin with, as.

A

D

A

So, for like I have one last new experiment, which is the donut ring. I think like it looks neat, but I always wonder uh so. It suffers the same problems as a pie. Chart right you can uh you can you can you can sort of compare like if the difference between them is big enough? You can see you can spot it easily right, but if it's subtle it becomes harder, and I think that's where a horizontal bar like a horizontal stacked thing is more powerful, but I guess it looks less fancy yeah.

A

um So that's my comment. The other one is, I think, oh I'm wondering if at some point we're going to move away from uh red green as uh success and error to uh blue and red and blue.

D

I'm actually looking into that I've been um doing some like exploration or sort of like oh, how did that happen? An audit of our colors and they're, not uh it's accessible.

D

These are our current success colors and they all look the same for two types of um color blindness and I spent probably an hour like moving my mouse around the red and the green and trying to find like a red, green combo.

D

That would pass the colorblind test and it's just impossible.

D

This is what our blue looks like, but then I feel like that's just an anti-pattern to so many applications and might be hard to learn that blue means it's up. I think it's a tricky problem.

C

I would actually think that we should actually talk with people who have these issues, because you sort of fooled yourself, sometimes because I've been doing this myself. You sort of look at these filter things and you see yeah, but they don't they look similar, but for a person who is used to seeing them. I think they can discern the difference there.

C

As long as there is a difference. Sometimes the colors just look exactly the same and that's an issue but- and you can also sort of play with how light and dark the colors are to sort of make them look different.

C

uh But for me usually, the solution with color is to have something more than color to differ differentiate them. So you have another thing that separates them. Just color.

G

D

All right, well, I've stolen a bunch of time, um even though I didn't add to the agenda so I'll stop presenting now, but thank you for the feedback I have to find my there. We.

A

Go um how much uh what time are we at uh yeah? I guess we're still pretty good in time. um uh I also don't really have anything to show yet I've programmed uh parts of the solution, so maybe I can. I can walk you a bit through what we're doing.

A

uh Let me share my.

A

A

So I'm on the sonification project, um we've identified some user stories, so one is as a developer, I'd like to hear if an error appears in the logs while I'm doing something else. Okay, so I think I thought that could be neat because you usually have a lot of log streams and um um one of the powerful things apart from accessibility is that you could sort of enable your your other senses as well to get to get a get alerted. If an error happens, and um so that's one thing, then the uh another one.

A

I think that is going to be more tricky. Is um we want to keep using the threshold mechanism within within how you how you define alerts in a time series panel on getting notified uh or audibly if that's being surpassed, or I guess also, then how, when it's kind of sort of come down um the more really more interesting one? Is this one?

A

I want to hear the ups and downs of a time series right, so if this could sort of completely be solidified and sort of have a melody being played, um that's something that we're working on. But it's not! It works like. I can get the data out and there is a sound engine.

I

A

I

A

Plugged together, so I think that's a bit sad, um but it's only wednesday um and uh then there's a more more intriguing one which is kind of similar, but it's based on heat map, where we could essentially take the heat map vertical slices as a spectrum and then could solidify those that could be interesting and then another more use case driven one, which is a network engineer, wants to hear if the pings sort of stop or if they change uh sort of a rhythm right.

A

So as you tweak your network um so far, uh yeah so working with costas. On this, we defined an api to make the solidifier and then that get that returns a function where I can just give it some numbers and then they're going to be turned into sounds, and so I'm working on the time series chart sonification, and you can't really see much within uh within grafana about this, because we, I kind of wanted to make it a central place. So uh let's hope this all.

A

Works so it would be just in here and then you can say.

A

Things like something's, not working.

A

I guess it's because I broke everything uh nope here here we go yeah so here then I can say: sonify right or I can say pa would have been really neat if ps would have still be available so actually like that we have a ton of shortcuts that are not documented. I noticed that uh now when I looked in the code, any sort of panel action is prefixed with a p, and then you can do another letter.

A

um I think we should have those somewhere anyway, so I just chose a for audio, so p a and then, if I do that, I actually get in the console here. I can see now the output of that time, series in sort of data and time pairs and that's going to be sent to a solidification module that crosstalk is working on. So that part is almost done and then the other thing is was the explorer story which is um uh in this live view here.

A

So if I do this, uh there's a lot of new lines coming in right and some of these have level. Let me just pause.

A

Some of these have level error right this one here, for example, and uh when sonify is on which I still have to tweak a bit here. This thing, um then it's also going to send the number essentially and the time to the solidification module and since log levels are also can be looked at as a letter. So, for example, debug would be, one info could be, two uh warning could be: three error could be four uh failure or critical could be five.

D

Yeah, so you you essentially.

A

Have a scale right like and those could be mapped to notes, and then you could also say right. I want everything above error right and then you get sort of different beeps trying to get sort of a beep for an error, but a beep for critical right, so that'll be cool um and that's what I have so far yeah any feedback.

E

I I just wanted to say like it would be like be totally insane if this ends up in with something that's actually usable, because I I love it.

A

Okay, cool yeah; no, I mean uh I'm, so I'm I'm fairly confident that this is gonna get into the product, one like at some way in some point in some shape. I just wonder if my hacky code is kind of made into the product or if, if someone needs to go and implement it properly,.

D

Is it um standard for applications to have sort of like a a code for what different things mean like you were explaining one two three four could match with different.

A

Logs, let me show you, let me share a different screen. Stop share uh the answer. The answer is yes uh there. It is.

D

Okay, I was just curious.

A

The developers have uh thought about this, and there is a sort of scale that you can see here. So it goes from. The problem is that different people have used different words and they sort of map to uh to this thing so uh info information. That's all info air error, like is this this. This is super annoying. There was. This is one logging library that said all log levels should have four letters and like four characters right, so they like crippled words uh yeah, that's just annoying, so debug is debug.

A

Error is not error or error. It's error with one r and crit yeah anyway, yeah, so there's a bunch of logging frameworks out there, and but they all map essentially to this other scale. Here right, like critical warning, error, info uh debug and trace.

D

um I think it might be. I don't know if this is even remotely possible, but it could be cool for people that really need the sonification for them to almost like customize. What different things mean like? Maybe there are defaults, but there is the ability to customize it for your own purposes.

A

Yep but yeah, I I totally agree in that. In that scale scenario, I also uh we have that little prototype here where, in our little dock.

A

There is here the logging alerts right, so there's like a sound on off, but then there's also this scale slider here where I can say all right. I want alerts over there, so I'm also making clear what are the scales uh like? What are the scale parts and um above which do? I want to be notified when a logline comes right, yeah, oh yeah, and then there's a another one which I'm I think is quite ambitious, but that would also be really cool because then it doesn't really.

A

Then it doesn't need sort of a time series in between. But if, uh if you look at an access log so like something that a web server outputs all the time and you won't anytime, someone makes a request- they usually have duration times in there and then, if you have a regular expression like this one here, like a little rule that can extract the thing. That's in this here you could then uh say: okay! Well, I wanna I want whatever comes in. I want this mapped onto this other range, where I say: zero.

D

Is obviously super fast.

A

So that's fine and uh like two thousand would be two would be two seconds so two thousand milliseconds. That would be a pretty long time for a request to be answered, and that would be like a pretty high or loud noise. And then anything is sort of in between um gets mapped to a sound and then so as the requests.

D

A

You get like baby anyway, that's.

A

It yeah all right.

E

D

Do that um that sound? um The thing you just did during the demo.

A

I mean so I don't know if we have time to have enough time, but I think we want to build a fake app that outputs logs in a certain range, so they could do this. One here right like this, is like this is the notes of funkytown right. So it's something like this.

A

A

Stop sharing, who is next, olaf or anyone else uh from um like a hackathon.

E

Project so uh olof can probably go first, but uh I think we have time then I can just show what we have come up with so far. All of can go first.

F

A

C

F

F

Is it this I think, oh my god, I think I have to.

F

I haven't allowed soon to be able to share my screen, uh so I guess I'll have to quit soon, but then maybe maybe tobias can go first. I guess because I have to restart this.

E

Okay- okay, sure I can. I can do that. Yeah.

C

Just everyone going to do the first presentation.

A

E

E

Okay, cool. Can everybody see everything.

E

Yep all righty right uh yeah, so we, uh this is not the correct.

E

Okay, let's hope that works in a reasonably quick fashion uh yeah. So basically, we wanted to um be able to like uh get jupiter notebooks to run in grafana and like the optimal solution, would be that you could like get data from like a panel or like a query and then just like process it in a in a jupyter notebook.

E

And like we ran into into some difficulties, the initial thought was like: oh, you can probably like it would be super cool. If we had like a document, we could, like just add a snapshot and then kind of describe it, uh and then you can maybe do some data handling with it, but uh turn up put to be super hard to like put either a snapshot or a panel outside of a dashboard.

E

uh So that did not work.

E

Oh there we go cool yeah, so uh we did something just a bit basic. uh Let's do a demo, so it basically just oh no.

E

Oh well, that's that I guess it's.

E

Oh. I think I have to.

A

Do that's annoying.

E

Oh well, we had a working notebook, at least, uh even though we had like issue of getting data into it. uh Great fun, reception, yeah, yeah, it's being uh the whole notebook, is being run inside an iframe which is kind of makes it super super tricky to like pass back and forth data.

E

um I don't think I might. I might not be able to get this to work actually.

E

E

Well, nope, it broke.

E

Yeah, I don't think I don't think I'm gonna be able to show anything.

E

It worked this morning. Okay, oh well, I'm gonna! Stop sharing see if I can get it to work.

A

Did you have a look at how vs code integrates jupyter notebooks because I think they they do.

E

uh Well, no not vs code, but we did look at some like alternative solutions so, like uh we have looked at, uh I think it's called observable.

A

Yep, I think he wanted to say observable hq.

B

A

All right, I think, it's time for one off.

F

Yeah: okay: let's see if this works now.

F

Can you see my screen yeah, no, okay cool, so the idea is that we have these log messages here that are kind of messy, and it's hard to read these. So I've added this kind of prettify switch that you can toggle to make them uh like more readable or like prettifying them.

F

uh So this is kind of the idea, but what I would like some feedback on is that sometimes we might not have. This is only possible when the log messages are in json and sometimes we might not have any json logs at all and then basically, this uh switch won't do anything.

F

So I wanted your feedback on if this kind of switch should be hidden or if it should just be, you know, disabled kind of or if there's any other way to to implement this uh kind of behavior uh and there's also a question of um people might want to always predify their logs for like any kind of data source.

F

So I was, I was thinking, maybe to add something in the settings of a data source to always predefine log messages, for example, and I will also like your feedback on if that's kind of a good idea or if, if there's any solution to that.

F

So yeah, if you have any feedback, just go ahead and share.

H

Maybe just for the sake of conversation, I would ask if users ever want to have not pretty sidewalks like if we shouldn't just do it defaultly um and like have that like what would be the reason? Why would they don't want to have this.

F

Yeah, that's actually good a good question. I don't really know. I guess if the like, the the log messages are like short enough, I guess there's no need for predicting them, but then also there's like this.

F

If it would always be prettified, then there's this kind of you have to opt out from that. If you don't want it and I guess it makes more sense to be able to opt in for it. I don't know.

J

Regarding this, like, I know the ecs specification for elasticsearch uh it's it's quite huge. I mean that a log uh document can have hundreds of uh fields that are really not related to the lock itself, sometimes so doing it. Like always doing, it means that basically, a person looking at elastic search logs will look at probably, I don't know like we'll, take three screens uh to only visualize a single log entry, so I'm not really sure about it.

J

F

In that case, you maybe wouldn't want it to always be pretty fine. I guess.

J

Yeah, at the same time, like even the default uh visualization is not going to be super helpful. In that case, yeah.

F

J

Will still have like a huge document anyway,.

B

um I really like the idea: uh it's it's a bit tricky. uh I I remember when we did something similar, but so we used to on a previous complete way way way back. We used to use like the elk stack so elastic log stash.

B

Then we basically had load processing like this, but we did it before we stored things in.

B

B

uh And correct me: if I drop, if I'm wrong, I think probably some of you guys have better knowledge about this, but there you always have like them, both the parsed document, and you also have the raw one. So you can always like toggle between the two.

B

So, for example, if you and and then we in that case, we used something like ingrock. I think it was called to basically write like parsing uh logic to log line, so it could be, um it could be json of the parts, but it could also be like just a uh a line of text and you just pick different parts of that line and said like okay. So this is the time stamp.

B

This is the message et cetera, so I'm not sure if it works to investigate, do something similar and but I I me, I think, as soon as you start to parse, to prettify the load lines, you also need to be able to see the raw one.

B

Just for that purpose, and just yeah.

F

G

I was wondering, like I know when you uh enable pretty fire, can we get like a different pain with the uh um you know, the uh in a structured data, but at the same time we can still like see the um you know. The original, like the original view, with the long log lines, or would that be like too clumsy uh to have them side by side.

F

um I don't know.

C

F

That, actually, that could be a solution, but the problem is that it might be a lot of information like if you could see both this like the raw message and the prettified.

F

uh It might take up a lot of space as well. I guess.

A

So, just to be clear, uh both are the raw message um just.

F

All right, yeah.

A

One is formatted.

F

Yeah yeah, it was just okay, so.

A

It's like slight distinction. Actually, we do remove nz codes there from the from the wrong message. So it's not exactly the wrong message, but um this is still like the prettification here is just there's no parsing happening per se, and this is just um there's just pretty finding it like indenting and formatting, um but to your point cheetah. If you prefer not to purify it, you can still expand the line and then you can see the parsed version, which looks a lot like the prettified version right.

A

So if you wanted to just do that, one right, you still have both like they're there. You actually have both.

A

But what helps a lot here, I think, is uh for people who don't want to do this on every row right because, like expansion, you need to do in every row. um I think there is benefit in having a predefined version, definitely available yeah and then, if they have huge documents yeah, then that's their problem.

A

Much we can do with uh well with loki. You actually can uh you can you can parse this as json? You can extract fields that you want to show you can like um you can jump into so, for example, like just you can jump into tags, only render the tags here. I think uh that's quite powerful, um but uh yeah, but I guess back to your whole question. I think um should this be on for everyone.

A

I would say: that's a that's a viewer preference, so someone, I think I would if it was up to me. I would store this in the local storage of the browser for the user, and then they can.

A

They can decide if they prefer using looking at this uh prettified or if they want to look at this and not verify it and then sort of store that preference by data source even right, so that yeah um that you don't because some data sources will more naturally have json data that looks sort of like this, and other data sources might just have um uh like yeah just other other structures. Here.

F

Do you think that sort of preference should be able to be toggled like or set here, or do you think it should be specified in like these settings for the data source or.

A

So so so I'm arguing for it being a view setting. So I think it makes a lot of like next two rep lines, I'm just I'm just slightly annoyed at the spacing of the label and the toggles, because it's a bit clearer, toggle left right, but- um and you had one more question which was is jason, the only structured, login format. uh No there's log format right, which gives you key values.

A

um And then so you have one one uh yeah.

D

A

One um one level of key values, as opposed to json, can be multi-level, multi-level steep.

A

So if you look at like grafana's lawns, anything in kafana outputs is formatted as with log format, which is a key value format, so we have a key equals a value and then those could be separated out onto individual rows when formatted yeah and then um to another comment here on the maybe prettify. Maybe it's not I mean that's not most ideal one. I would probably call it format.

A

Formats, format, structured log lines, I think, would be sounds too long, but I think that's most accurate yeah, so maybe pretty fire is okay um yeah, because we can only really format the structured laws.

F

Yeah I'll have a brainstorm session later and try to come up with something else. I guess I was also wondering about like when we don't have in the case in this case, when we're only looking for json logs.

F

Sometimes we don't have any json logs and then this switch won't do anything. Do you think we should just hide this, or should it be disabled or.

H

So I I was thinking about it uh now and I'm wondering so. I don't like when staff are disabled or hidden because then sometimes it creates this feeling of confusion. Then I don't know. Why is that? And what should I do to make it not uh disabled, uh but I'm just thinking like if you would pretty paid and there won't be any logs in lock format on or json format. Maybe we could add um like meta info message or some like info for user. Like hey your logs, like you, don't have formatted or structured logs.

D

H

Happened um like like just uh my thoughts would be to have it like. Similarly, as you have wine limit so just like out, there inform we usually add there some meta information, so maybe we could add something there or I don't know, maybe someone else's idea, but I I personally don't like disabled stuff.

H

When I don't know, why is that.

G

B

Thanks now, when I got some more context, I think what what I would.

B

Like and it would be kind of cool to, instead of have like a pretty fine, it would be nice to be able to select what fields that should be displayed in in, like the collapsed.

F

Mode so like only only showing like message and time stamp, for example,.

B

Yeah, so basically I could like.

H

B

Sorry go ahead.

H

Just tell you you can actually.

F

I think you can use these to just choose right, so I'm only looking at the message right now and then I can also get message and timestamp here, but then you would. Okay, no.

B

Then I'm already happy sorry, nice, but that's just me.

B

I just realized: I need to use these features way more.

F

Yeah they're pretty pretty neat.

F

Okay, so does anyone else have any other feedback that they'd like to share.

J

My only thing like uh something uh for me like having the the prettify option is actually nice. uh The only thing is that uh I think, like it's, not the first, the first time we we had this speed back on on the options uh here, it's that it started to grow a bit uh and really I'm having some troubles, understanding which toggle is uh referring to the predefined, which one is referring to wrap lines. I don't like the spacing, is not optimal.

F

It would be nice to have some sort of dividers.

F

A

Yep yep, I think.

D

There will be something for the design.

F

F

Well, thank you for the feedback.

A

Cool thanks for presenting olaf and everyone else who presented um anything how we can improve the feedback sessions.

A

Any meta feedback.

A

What did you think mary first time.

I

I think, oh sorry, my video's been a little weird. I think it was interesting to see the products and actually, I think uh something that would have been interesting for your presentation would have been to actually hear the sounds. I was going to ask you. um I know you did the little blah blah blah blah blah, but to actually kind of like hear what some of the sounds were like that have been really cool. I think it sounds really awesome. So thank you.

A

A

We haven't brought the codes together yet yeah. That's the sad side.

E

Make sure the stuff you're supposed to present works.

A

My self-reflection uh teddy you're also muted.

D

Oh, um I I noticed that um everyone today like started off by saying what kind of feedback they're looking for, and I feel like. That's really helpful.

D

Just going forward for people to.

D

A

All right um thanks we're exactly done after an hour, um then um have an awesome wednesday. The rest of the week.

I

I