Open Telemetry Uncategorized, 25 May 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: 2022-05-25 meeting

Description

OpenTelemetry PHP SIG Meeting

A

A

A

B

C

Good morning,.

C

D

Hello, everyone good day.

C

C

We could start in a couple minutes.

C

Oh captain, my captain.

C

This coffee must go directly into my bloodstream. That'd be really great.

B

C

Not yet they're working on it.

B

I'm coffee, I'm sure.

C

Isn't that just a caffeine pill.

B

Probably yeah: well, I don't think it's super healthy shooting caffeine in the bloodstream.

C

Yeah, I don't think I think you're right timo, but I'm willing to try.

C

Especially with a young child, the no sleep things that is real.

C

But she's gonna start working on open telemetry, php.

B

Daughter, yeah she's.

C

Yeah, I just figured she's two two years old, now she's it's about time for her to start writing some code.

B

Yeah yeah she can start with writing some tests.

C

I'm not going to make any off color jokes. Thank you I'll, be very happy to share my screen with discussing pr's and issues, so people can follow along thanks for the suggestion.

A

A

Is 805. I think we can get discussing I'll share my screen real, quick.

C

Y'all see my github window.

A

C

B

C

Yeah dark mode. Thankfully, that's right! Good morning, everyone good day, everyone, let's get started with taking a look at our open, prs and open bugs and we'll go through discussion points. It doesn't seem like anybody added any to today's agenda, but I think we have enough people here, we'll build a quorum.

C

B

And you wrote my family name wrong.

C

B

B

The n has to go after the h.

C

I'm sorry one more time.

B

The n yeah yes.

C

C

All right so look at the pull requests that we have open the open, telemetry, php repo, uh this creating events and listeners that one's been working on forever. So we saw your uh your draft for the otlp grpc export implementation, and thank you for working on that. It's looking really really nice. uh I think brett and tino made some comments on this one.

D

Because I just wanted to make sure the approach is approved by all the maintainers and after that I'll add the test cases. So.

A

D

Can see the brett has uh commented that you know uh suggested a different approach for the retries so I'll try to implement that.

C

Cool, that's good yeah. I think any time you uh just in general. Any time you can avoid a sleep at any cost is good unless you, unless the logic requires it, because it's a waste of time, but I'm sure that I, like brett's approach.

C

You know it's that's a very difficult concept to start working on, because a lot of times there's a lot of moving parts and you know work together, but I I really appreciate your work on this. It looks really good. I'm excited for you to see.

B

A

C

A

um Sure, thank you.

D

B

D

B

Okay, well, the problem is, maybe we could, in the future, solve this a little bit more elegantly with fibers, so that you could get back the context. You know you don't have to sleep um but yeah for now that, since we have to support hp 7.4, it's not possible and the only thing, because so this may maybe not message the comments, because I didn't look deeper into it. Yet um what we have to have is that um the default is basically who we try to records, because.

B

I think we discussed this in this initial ticket because we have to consider normally we have a execution stack in apache or nginx, or something like this, and we cannot make um the whole application on the execution rate, because some http request of has to wait. So this actually has to be configured for people to have more than one retry, or at least. However, you try.

B

Do you know what I mean.

D

B

Okay, well then, I can, when I take a look into the um ticket ix, can exchange further well in general. The problem is when we have a request, and normally in the classic execution stack. It's one process, one thread, and then it works goes through all the php um processes to spin up the to spin up the process and basically then some point the poster gets close and that's when usually the last headers and everything is sent to the browser.

B

If we wait longer um for http requests to to go out, um then it could either well in the best case, it's just making the browser and the makes it spin. So it doesn't. It's not really completed the process. In the worst case, we will run into next execution time issue so because there's not so much time configured to be yeah which can be used to execute on the process.

B

Because of that it should be. We should be have ability to configure the number from choice and it should be actually zero for default.

D

uh Actually, the retries are configurable, uh I may not have provided proper. You know, example a document, but uh the retry is there is a interface and there is a you know: implementation for the data, so you can, while you can, when you use the child class you at that time, you can configure the you know: retry mechanism, okay,.

B

Okay, great, like I said I didn't look into it. It was just a discussion with some.

B

D

If you can see this on the line 73 for exporter php, I can specify the this.retry policy and uh provided the get default, but instead of by default, you can provide all the parameters that are configurable.

D

And you can configure all the parameters individually itself, so not just no. We don't have to provide everything in the constructor itself, but you can provide all the parameters separately. Yeah. All this max attempt initial pack off max back off everything all the five six parameters you can specify them separately, as well as in construct.

D

A

D

Sorry, I just said it looks good okay.

C

All right, uh kashan, you had this at a b3 multiproctator is this: um is this ready for a review.

B

C

Okay, we, I will review it after this uh I'll review. It today.

B

Okay, I don't have to.

C

Yeah, don't worry, yeah, don't worry about it I'll. Do it I'll.

B

Do it, I already took a short look, so it's it's! Oh.

A

B

I achieved it: um lots of tests.

C

That's good everybody likes tests.

C

Let's see, okay so in the contrib repo uh teemo. Thank you so much for all your work with the package. Well, that's been really uh really fun to watch.

B

Yeah, oh sorry, go ahead. I.

C

Was gonna say I know you have some comments about this.

B

Some some what.

C

Comments about the package, but.

B

Yeah, so this was.

B

If everybody is not familiar now, we have basically split all the packages like we planned for a few months ago um into separate composer packages which are also published on packages.org. So that's basically.

B

For now the two components or packages we have in the contract repository and then I think it's six or seven for the um for the main repository, so the api context, um sdk conventions and then the two generated files. We have with thrift, thrift, jaeger and what's the other one otp and the prototype files, and now they can actually be.

B

Required or dependent on.

B

Yeah, if you just go the first one, the first one, just click on open telemetry, it should show all the no just in the title.

C

Oh, you mean just the title.

B

And then it should show you. Yes, all the packages which have been downloaded millions of times already.

C

Not quite not quite millions, but the open telemetry one has got been downloaded over 60 thousand times. That's kind of cool yeah.

B

I mean this is the main package, um so now we can require the synthetic stuff separately and I'm not really 100 done with it. Yet so I didn't close the respective ticket yet because I would have to add some tests um validating that the composer files are correct and then actually everything can be installed.

B

So that's a little bit of work on my case and then some cleanup, because not everything's still correct there, but at the end of the day this is done and there's been more manual work than I would have liked, but I will add the automation for the future, because I don't want to repeat all these steps for everyone. The package we have because at the moment has been quite a little bit tedious, but I will work in the future.

B

uh The additional thing is um this is going especially to bob and for this validation there's not really something I can. I can reuse some commands, but not really not an easy way. So I decided to create a little.

B

Dev tools package, which I haven't published yet which we can put on or an open timer to php repository in the organization. So that's just a little wrapper.

B

I don't want to put it in the main repo because it will be a separate package that will be then used by just by the main repo and the country, people as well. So.

C

Yeah, that makes sense very cool.

B

Yeah then maybe we can add some other stuff in the future for whatever testing. When we work on the etv tests and things like that, then you can just create some commands and yeah I'll run them from ci or wherever.

C

Yeah we at the company that I work at we filed the github pattern with slash with that script, slash whatever I don't know. If you're familiar with that pattern, I can find the link for it, but I like your idea better.

C

Now, of course, you got, I don't want to go too far in the rabbit hole.

C

Yeah they came up with this yeah here it is there's uh they came up with this pattern. You have a folder called slash script, slash whatever, and all these end up being executable files. These are probably all just bash, but it's just real a really nice way to be able to just go dot, slash, script, dot, slash, bootstrap, and so on so forth.

C

We could do that with no repo, but I like your idea of creating a package to handle it all better.

B

Yeah, because we need to have them on two repositories, so right, that's yep agreed um initially, I thought to put this stuff over in the main repo buckets. I think it's better to have this one separately.

C

A little bit easier to maintain yeah, wonderful, all right. Those were the open, pull requests. Yeah. Oh.

B

And another thing: no, no, um regarding, if you go back, you see two packages yeah to this one. um You can actually see the two ones we talked about some weeks ago of separating and there's actually a ticket for it. So that's the gen otlp protobot and gen gigathrift.

B

That's the reason why I use this prefix gen, because we have a ticket to separate the generator to put them basically in separate repositories, because they have their own versioning, which depends on the downstream, whatever the thrift downstream or the reporter.

B

And would be better to follow those uh version types and would be we could actually create some water generation and wouldn't don't have to take much care of it and on the other hand, um the packages could be reduced to um by other projects which may or may not move and yeah. For those, then we will probably then drop the gem um prefix. So that's basically just it when we once without the prefix will just be the raw package.

C

Very good, uh alongside that, I'm going I'll publish a release today because we haven't published one in a month.

C

Do we want to get do we want to get those on this? Should we get the um the main repo and the contributor on the same release, cadence.

B

C

Are we okay with publishing them at will still.

B

um I yeah it would be nice to have the same release games, um but on the other hand, what we talked some weeks ago is that we actually want to follow, um but um the versioning the version number or follow with the version of the contract repository well, we want to have the same version number on uh the main repository and the contributor repository, but we will need some tests for that, so that we actually have a check um when we have a pull request in main if it breaks something on the conflict repository not not like a check that would prevent, um prevent, merging or something like this, but just an indicator that there has to be something that has to be fixed and that's also, the reason why we put in this compatibility layer to uh make this sure.

B

So I yeah, but I think um it's easier to deal with all this.

B

If every package has the same versioning number, then we know if it has the same version that should work together with all right. The same version right.

C

Right, yeah, that's that's! I think that's so I I think that that's a good pro for keeping the same version number, the negative for keeping that same version number would be. Sometimes if you have disparate version numbers, it can be easier to troubleshoot when somebody says version 0.0.10 right now. They know that means the main repo, not the contrib repo, but I don't think that that's enough of a poll to um skip that. I I'm in your camp here team. I think that they should be on the same route.

C

They should have the same release number on the same release. Cadence.

C

We could just skip a couple versions of the contrib repo. If we wanted to just go straight to 0.0.11, it's not really that important.

B

Yeah, we don't have to do it now, because there's still still some things to fix to make sure that all this works.

C

That almost feels like first major version, will release 1.0 with both at the same time,.

B

Yeah, that's definitely, but um when we get closer to beta or even other release diamond ruby, we won't uh break things as often as we used to um in the past. So for the country.

B

Probably hopefully I don't have to fix stuff so often.

C

B

Especially not for companies.

B

Three letters.

C

Oh yeah right it rhymes with the major, maybe.

C

uh All right, so that was all the things that we had in the repo and on the agenda. Does anybody else have anything that they would like to discuss today.

B

C

B

Don't think I can talk for another hour, but um no, I don't think there should be anything more.

C

All right sounds good all right if, uh if anyone has any additional questions or concerns, please raise some in the channel and we'll uh catch y'all next week, thanks for attending everyone.

B

Bye, everyone have a nice week. I know.

D

Thank you. Thank you. Everyone. Thank you.

A