Open Telemetry Uncategorized, 5 Apr 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: 2023-04-05 meeting

Description

OpenTelemetry PHP SIG Meeting

A

A

You're sneaky my uh we're teaching my daughter's sneak attack right now and she's loving it. So it's.

B

A

That he's like you're, trying to sneak in.

C

uh I love a good sneak attack. You and me: does it involve tickling? Oh.

A

No, it's just like guess what I made it home from school, but I guess I should probably incorporate some tickling there. That's a you know, that's a really good idea. Thanks for the advice also I would love for you to know that I've started using good day as my greeting for a lot of different things. Heavy.

C

A

I decided yeah it's time. You know we talk so much about gender neutrality. We don't talk a whole lot about time. Neutrality.

A

Maybe there's some: maybe there are some time is up there. I it is for that. Is it actually is kind of a fun solution to a problem that is kind of annoying I? Really, it's really annoying when people say good morning when they're, when you work for a West, Coast company like it's not morning for me it's afternoon, Yes.

C

A

Y'all Brits has solved the problem already.

A

C

A

Said experts, I guess I should say yeah.

B

D

C

D

Welcome Cedric.

E

C

To me, but I check in.

C

All right don't mind us. We just uh talk rubbish for a few minutes until uh until some people show up.

D

C

A

Right, you gotta be a little bit more explicit about that. We tuck rubbish until everybody shows up, and then we talk.

B

A

We took different rubbing sheets yeah different rubbish. That's it all right, I! Suppose we can get going because I don't expect anybody else to be here today.

C

A

Do it in one moment let.

C

A

Get rid of all my company secrets and share my screen. That's actually an important topic today that I put into our agenda is secrecy.

C

Is I saw that come up here now, maintain it somewhere yeah.

A

I don't want to first I. Don't want to get ahead of myself all right. Can you all see my agenda? Yes, yes,.

B

B

A

It's the open, pull requests for the main repo. You got two from you Brad. You want to talk to these.

C

Yes, yes, I do um but yeah, let's start at the bottom. um Oh, this is just a PHP doc verification, because I had some questions where I received some questions from somebody who's trying to use it, and just it wasn't clear that um two uses nanoseconds a time so um yeah, that's a that's a really trivial.

B

One um convenience thank.

C

B

A

I saw this when I have not yet reviewed this one. This is.

B

A

C

More, that's fine, so so I've started writing a a Handler for monologue to integrate into um up to entry money, as described in the sort of log specification, um and what actually got to writing it. I see found a few inconsistencies with how the specs that it should work out and how I did it originally implemented. The logging API so I've got some questions to get away called the maintainers list um which I'd like to to resolve first, but I think that this is I, think I'm happy with with how it works.

C

I just want to make sure that a child. You know none of the the other maintainers are still having issues with the web. I've done it.

A

I think, at least from my perspective, Jack, the guy that was responding to an Administration, that's his name right, Jack, Jack,.

C

A

Yeah uh he's the right person, because I think he was pretty integral in writing the log spec and he talks about it. Frequently he's the one who gives updates for the long spec you could.

C

A

Yes, he definitely is the right person to answer them.

C

And the right guy yep so so I'm pretty happy with that, and it works well with what I look um implementation that I've sort of got a branch going for um just get yep, just like so some sort of green checks from somebody in maintain this before I.

C

You know before it gets reviewed here.

B

Okay, I will uh I'll argue that later today, in trip has one.

A

So I got uh can I need to prove this. It looks like this person just doesn't. Oh yes,.

D

It seems that this person stopped working on that and uh we have to uh apply this comment that uh or suggestions that that Brett suggested so I don't know. I can take uh take over that so awesome. Thank you. Yeah.

C

Maybe just um well you're in there, but you're just pinging planning to work on it.

B

B

No instrumentation PR's, no new discussions.

B

New New Circle overflow questions.

B

Something all right um open: let's take a look at open, bugs.

A

Oh yes, project this one, this one's the one that we were talking about with uh it's like I, think you said you were gonna. Take a look at this today.

C

A

Is the sick boat? This is a different one.

C

A

C

uh This is a complex, so I've had a couple of a couple of good. Well, you know yeah, my my checking of the um telemetry package that we abandoned.

A

Yeah, that's on your that's on your agenda.

C

Right, yeah, yeah.

A

We can talk about it.

C

uh Yeah sure, okay um well I mean it's it's just that um a few people were, let's see if people had, um you know like composer, lock files using or they're just their codes and seeing sort of somewhere quite old um Alpha releases, um yeah they're, experiencing issues for not being able to install to build failures, um so I sit throughout races in seek and talk about it.

C

um So I guess the question is: should we do something about it? Provide sort of this really old um versions somewhere on packages for people who are still using very old Alpha versions or, and is it enough of a? Is there enough demand for that to Warrant the extra work for us, I suppose, is the question I'm, certainly not looking for more work to do so, it depends how much maintenance involved.

D

Yeah and the same problem we have with this Symphony package, uh so it stopped. You know working after this change, so now it requires big update. I would.

C

Say yeah: well, it's now more broken, but it was already very broken before um you know it needed help. But yes, yes, I, agree.

E

It's actually the reason why I'm here, oh.

B

C

uh Do you want to talk to her.

E

Just briefly um so I started the effort to like create a bundle, but then I had to abandon it um and I was picking it up again. A few weeks back and um I saw that there was like this drift between the dependencies.

E

So uh when do you think, is the right point in time to um to adjust what we, what we have in code with uh with the current dependency structure,.

C

uh Oh I think right now, um there's there's no reason to to hold off, um um maybe talk to us on slack a there's, a Direction, but um in terms of like, if all of the core of that code is, is deep into beta now and approaching ga equality.

C

um So one of the issues with the the symphony bundle was it was using sort of the whole um or telemetry package, I think which was basically our entire whatever.

C

um So, if you were to break that out into probably just the API, oh actually, no because it uses SDK and exporters Etc, um so so just swap that out for individual packages, so the API, the SDK, the open, torture, exporter, Etc um and use those because it using all those individually could be the same thing as using the old sort of kitchen sink better package.

E

Do you think it's important at this point that we recognize whether the extension is installed and do some other instrumentation or on I mean there's like this disconnect right? You would end up with uh with two two spans for the same thing, essentially theories that.

C

Is a very, very good question, um so obviously, this package predates Auto instrumentation by by quite a long time, um and that is that's very good question. It probably needs to be answered by someone with um sort of David Symphony knowledge like what what do you specifically use respect.

E

C

E

I could try to answer, but I was sort of sticking for for guidance. How were you folks I mean you were you are the primary authors and uh like what is your idea about the extension? Do you think it's like a sort of exclusive thing that you don't have any instrumentation in code anymore or um do we want to support the hybrid approach? I can certainly wiggle my way through and it you know, certainly possible to check for class existence or extension loaded or whatever to like circumvent.

E

But it's more about the question about the general approach. Let's be honest,.

D

That is something that I I was also was thinking about, because uh now the question is: if we really need this um kind of manual instrumentation as we have this uh Auto instrumentation, so we can generally rewrite all this code that we have in this Symphony bundle and to make it auto instrumentation.

D

um You know kind.

E

I I would like to disagree, but I don't want to open that kind of words right now I. We are not through the agenda yet.

A

I agree: I think that we always have to allow for manual instrumentation when necessary. I think that's an important part of being able to wrap a spin around whatever you want to.

C

Yeah yeah I I'd agree with that and and we can have both as well um yeah yeah the country is certainly has a has a lower barrier to entry. We can have things that are pretty incompatible with each other, provide a new document that you probably don't want to use the manual instrumentation and a future Auto instrumentation um yeah I. Guess we like Auto instrumentation, because it's it's quite efficient and it's a nice separation of your code from the instrumentation code.

C

But that's um that doesn't mean that that everything has to ex has to be written with auto instrumentation in mind.

D

C

D

True, but from the user's perspective this uh this uh manual, instrumentation, that we have in contrib is a kind of Auto instrumentation I I mean from the user perspective right because uh still user doesn't need to do any changes in his code. That's that's how I uh see it.

C

That is true, that is true, and it's probably doing almost simply way.

E

Yeah, so, okay, the the symphony way, is definitely to use dependency injection to uh like modify the state of the application so that it would record Telemetry and not do it from the outside um I I guess um my my point in regards to relying completely on the extension.

E

um This is, in my opinion, not a good idea, as you cannot roll out an uh an extension is especially if it's not pre-compile compiled available to to like any environment, and if we fully relied on the extension to do the instrumentation work, then um we would kind of exclude the people that are using more classical hosting approaches.

D

E

D

C

uh Yeah I agree. So look, it sounds like what we're saying is that they they're certainly a requirement or a very valid use case for the uh Symphony bundle as it exists now um once once it is working um and then you know it's a good option.

C

We may in the future, also Auto instrument um Symphony framework, but those two things can certainly coexist and the existing bundle doesn't need to use Auto instrumentation. It doesn't need to morph into Auto instrumentation, as we understand it with the extension.

D

But does it mean that we would like to follow the same approach for other Frameworks, so I mean to have let's say uh a kind of manual instrumentation and the automatic instrumentation, with the with the extension.

C

I mean a surprise: if someone cares enough to write one I mean we do have that the laravel, although it's posted outside of the open Telemetry project, which is the the one that you've been sort of trying to re-architect with auto instrumentation, which is which is a good option for those who can't or don't want to install an extension.

D

Because now, as we are talking about it, I think that we can just get this library from from Scott blanked this Lara verv1 and just you know, merge it I see this may be. Small changes will be needed, but we can add it in the same way as the symphony. One.

D

C

Think we would happily accepted.

C

uh So Cedric do have. We helped you at all to understand absolutely what you what you want to do next.

E

Yes, yes, absolutely you. Thank you so much excellent. Thank.

C

You um so if we just switch back to um so, let's talk about the the library but um I guess that is about um dude. Do we need to should we provide the old meta package in some form? For you know what is probably a small number of people but but I'm getting less non-zero number of people who were relying on it or do we just encourage them to upgrade? Because and then you know, the number of people affected is already low and heading towards zero.

A

I think that um I have two different opinions on these, but one is itching a lot stronger than the other one. My opinions are. We should try and support people. However, we can to use our library, that's one, but that's the less that you won the real that you want is like this is where this is not a generally available product. Yet we, like we've, been very explicit that our contracts may change and those people should just update the library. It's not that hard.

C

Yeah yeah well and the person who sort of was um I guess the most vocal about having it fixed, has I think since just upgraded anyway, you know within the yeah, maybe.

A

It's interesting yeah, that's why I was saying. Maybe we should raise an issue like or add some documentation like if you fall into this trap, just upgrade to the latest version or maybe create an issue that says um you know trackings. If you need this, but that just opens a can of worms too I think we just my opinion. Is we just do like no? No backwards compatibility use the new one yeah.

C

I mean there is a pretty good work around, which is just adding the mono repo that has a packaging composer and that it does work um it just um in in this particular user's case. You, like their company policy, is don't do that because they're not allowed to connect or including you know for.

C

um Yes, but I think for the average for the average hacker. That's a pretty good work around.

A

Yeah I think we have bigger fish to fry than trying to maintain a package for one third tenths of people like I'm, trying to maintain a work around for ones to tens of people.

C

Yeah yeah I probably agree with that.

A

And you could say: if you want to do it open source, do you do it.

C

A

Yes, I think I think your talent can be used other other places, rather than maintaining that backwards. Compatibility.

C

Well am I limited time is, is the thing, but then how much.

A

We shouldn't make it more limited for goofy stuff like that. Anyway, all right, uh that's my opinion. Let's.

C

Move on that's good, uh sorry, we're still getting three bugs or are we up to each into now.

A

We were going through bugs, but we, uh but we.

B

Got sidetracked by.

A

That, okay, so this uh PHP fpn Cedric, are necessary. um I think you said you were planning on working on this.

D

Yeah yeah, so I was busy this week with other stuff but I'm. Currently, returning to this sick faults and I plan to to focus on that.

B

A

Cool, thank you very much that I've seen that in a couple of places, so I really appreciate you working on that. That's great.

B

With a couple old ones,.

A

Anybody want to talk about any of the rest of these, or should we move on.

C

I'm happy to move on.

B

A

Right tackle access, yeah.

C

A

Two exclamation points that might.

C

Not be enough, how do I think it's enough? It was hot that was hot, well, tedious, anyway, uh yeah just just pat on the bank for us that uh we got it done in a deep name. It's only for everyone. So that's.

A

Great uh documentation tax is merged. Thank you. Brett.

C

A

C

Yeah, just just to mention that there's there's still once more documentation to to be tied it up and added on and telemetry.io.

C

um But yeah most of our documentation now is just links into hotel.io, so um I just want to call out a little, probably just nuke. Most of that and just have a top level read the talks go here and an implementary as the primary source of documentation, which is what sort of the maintainers. What I'm? Sorry that someone can't remember! That's what the powers that be would like.

A

Yes, the the council I think. Yes, that's true for I'm glad I'm very glad to hear this for a couple reasons. One I think that it's good that we're standardizing that uh on open Telemetry, I o I, think that's not what a lot of other uh sigs are doing and I think that it's good to fall in line there. I think the other thing is that GitHub doesn't get indexed by search engines very well. So.

D

A

Haven't uh having its own Standalone website is great and I agree with you that I think that it'll be really smart to have one centralized place, because I could see us having uh read me sprawl between all of our repositories, so it'll be nice to have one centralized place for that. I do Envision. Eventually in uh yeah we could I was thinking. We probably should also change this here. The open so I'm not sure I'll link to the PHP link.

B

Maybe two: yes,.

C

Yes, that's a good idea, it's under instrumentation. Yes, thank you.

A

Too early in the morning, for you.

C

Yes, I would agree with that. I think that's where it should go.

B

If we could just change this real quick.

B

There we go: okay, wonderful.

D

All right, I'm done talking.

A

D

A

With your agenda at items that doesn't.

C

Necessarily and.

A

Then you're done talking so I have the next two gender items. I talked with the Sig maintainers this week in the maintenance meeting. That's on Monday and we talked about opening a full request in the community review for General availability quote unquote: checkup as they as we so lovingly name. It I'm planning on doing that this week and then I think Carlos is one of the TC members. The test committee or again I can never remember the acronyms he's going to he's planning on reviewing our repository and probably giving us some pretty meaningful feedback.

A

So that will be nice.

B

C

There be any sort of interview Style, you know we can't consume and booking through and answer questions or I.

A

I think I think it's intentionally him reading through it because I as I understand it. They try and do that review very independent of the maintainers, because it's easy to drive that conversation from the perspective of the maintainers rather than the perspective of the reviewer but I'm sure he'll I have no doubt that they'll be asking us questions. If they don't I would consider that review a failure.

C

A

My next agenda item is cat GPT, who I was in the maintainers meeting, so the maintainers a couple of maintainers and a couple other people in the community brought up a pretty good point. They were talking about being able to use, uh chat, GPT and deep fakes and some of these other tools to impersonate a an open, telemetry maintainer.

A

So they were talking about changing the retention policy for the maintainer meetings, like the one we're on now, they're all recorded to YouTube and saved for Infinity time. So they were talking about creating a retention policy there and they're. Also talking about requirement of video and requirement of discussion and blah blah blah. But the tool I didn't read is the retention policy for videos may change, and they nobody really has a great solution for this.

A

A lot of the maintainers were joking that you should just leave your camera off, and it's also important to know that if you have other online presence, it wouldn't just be limited to a sick meeting. So just wanted to let you on that I thought that was kind of interesting and kind of funny. At the same time, so I figured I'd bring it up in this meeting.

C

A

C

C

A

Okay, yeah and we're Progressive we're Progressive as it is by not using camera on our Sig meeting. So.

C

A

The last the floor is yours for low level instrumentations.

E

Hey that's cool, so I was maintaining of a vendor-specific tracing extension before um and um we want at the time the strategy was to use the lowest possible Primitives um to instrument so that the the reach and impact would be the biggest.

E

um So my question is: has this group thought about instrumenting um like runtime internals like I, built-in extension functionality or even core functions like file git contents, um the request in it event to create the entry span for the sorry, the service plan in Hawaii lingo? This is something that that you that you have considered and that do you have an opinion um about it.

D

So we are instrumenting already some of these internal functions, for instance Carl and some functions on uh some. I o functions for for uh file operations.

E

I missed that sorry.

D

So you can look at the country report, the reset I o package and we are hooking in the few few operations.

C

D

E

Very noisy, though so.

C

You know turn them on at your peril, depending on what your application does um as far as r in it and some of those internals, um it would have to be sort of an extension to the extension um because we're using the Observer API that came in with php8 and that allows us to observe or attach observers to PHP functions uh and those can be built-ins like curl and forget contents or or anything any use of land code.

C

You know files, sorry functions for you know, class methods, um so I think aside from priorities, uh we can do most of those things, um but I know no I. Don't think we've thought about um sort of diplomatically starting traces that early and I think one of the all of the walkers for doing that might be that our sdks are configured in code um and currently that happens just as part of well either defined by the user or as part of composer, auto loading.

C

If, if we're using a sort of SDK auto loading feature, so the SDK might actually be available too late for something. As early as orange as I understand that one.

E

No, that's fair, absolutely, um and so um there are other examples of this out there, but um basically we retroactively mangled the um the service plan. As soon as we were able to like recognize intrinsics um but but I guess, that's kind of a No-No in in Optometry anyways, as fans are considered uh kind of immutable right. You cannot change too many uh things about them.

C

You're right there's only the name, it's about the only thing you're all typically will help. The API allows you to change after the fact and there's also less of a distinction between a span but we're saved another way. Open Telemetry doesn't have the concept of a transaction that a lot of other PPM teams. Implementations do um and it's a lot harder to find what it was. The root.

B

C

Is about equivalent to a transaction, but but it's not special when you can't can't just get a handle on it and manipulate it like you do in, like you can in other APN products.

E

E

I think that answered my my question sufficiently. um I just added another item on the flow. um How do people feel about adding a trace response propagator so that the server would be able to communicate context to, for example, a JavaScript client without having to resort to like a meta tag in an HTML.

E

That will basically mean injecting another. The current Trace context as a new header onto almost any response, but look that would contain the active and context.

C

So is this operating in the opposite direction to the w3c trace context, headers.

E

uh Yeah there is a there is a sibling specification, it's for it's in draft mode for years and years, um and um it's basically about propagating the transparent to the client. Yes,.

C

um That sounds like a specification problem before before it's an implementation problem or, for our sake, um I think, I think that the way to tackle that would be to bring it up in the specification seek and get some support there and get it sort of added to the opportunity specification, uh at which point we would be I'm sure very supportive of implementing it uh Bob. Does that sound like 30 the right way to to.

A

Approach this that sounds exactly correct to me.

E

Yeah, so would you would you be supportive of including it in the Contra package? um I I see that other uh six have accepted those, for example, Ruby, Python and.

C

E

C

Assuming that they that it's part of the specification, it would be, you know, I- think it would live in the core repository um I think it will be part of the official API and implement it in sdks.

C

um Is that is that the question you're asking or or is the question? Would we accept it like as a temporary I.

E

Would I would like to start experimenting with it.

D

E

So full transparency I'm from grafana labs and we are creating client-side instrumentation um and um since my home is in PHP um I. No, that's that's my my sort of playground and I wanted to ask. Would you be supportive of including this kind of experimental propagator in the.

A

E

Package of which you see as something that's strictly um outside of of any sick maintained, Repository.

D

um Yeah, but is this part of the spec already because you said that the other languages have this feature.

E

Some some six have accepted Trace response propagators. Yes, yes, uh I, don't think it's in the spec. Okay.

D

A

That's that's kind of frustrating right. That's the whole point it's specification, but if.

D

B

Languages have accepted.

D

A

Languages have accepted it, then we will accept it. Yeah.

C

I would agree with that yeah. Definitely, if there's precedence, then absolutely.

A

C

A

I I know from experience this. uh The specifications Stig uh loses races to snails frequently, so that could be uh part of the problem.

E

B

C

E

Also a good point in time to propose it as the spec change um phasing it on the prior art in like many sick projects. I could also turn that around yeah.

A

Like hey, hey, y'all, this isn't in this, this isn't in the spec yet. But these.

C

Languages, yes,.

A

Can we just make it consistent so that there's not confusion, then I understand why you have a vested interest in that Cedric. So.

C

Yeah but yes, um an answer to your question, I think based on um based on you know, prior art and other seeks. Having already done this, we would certainly accept it in contrary, um but I still think it's a worthwhile exercise to um uh try and get it implemented into the specification.

C

um Sometimes people just need to be reminded of um I. Think of interested someone to provide some energy um because then, if it becomes part of the specification, um we can put it into the core repository which sounds like where it would be better housed.

C

E

Thanks a lot for for the discussion, much appreciated.

C

A

You for asking in person in uh and meeting in person all right. Thank you for coming to the Sig meeting. That's always it's always helpful to do that then talk on the GitHub issue for two weeks.

E

Yes, absolutely I'm intending to do that then more regularly. Excellent.

C

um Yeah well, when you, when you do get to you, know that being ready, you've got a PR in um we'll be happy to help you just with the last little bits of getting it into its own um uh packaging packages, to do all the git splitting and stuff that we do. But it's a little bit of maintainer requirement there. But um pretty.

B

Pretty trivial stuff.

E

C

All right are we done.

A

I think so this is one of the longest six segments we've had in a while a.

C

B

C

A

All right, we'll catch you all on the internet, all.

C

Right thanks for your time, everybody.

D

Good day, good day, foreign.