Open Telemetry Uncategorized, 21 Nov 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: 2022-11-21 meeting

Description

cncf-opentelemetry meeting-2's Personal Meeting Room

A

Hey Mark ISL.

B

Say: hi, hello. Everyone thanks for joining.

B

Let me share my screen.

C

B

Yeah so people add the points in the agenda. uh I have added some of that stuff, which we can discuss, at least for the specs changes.

B

The font is visible.

D

B

It's not small.

A

It's okay, okay,.

B

So the first one is basically the capturing the policy regarding non-exposure of generated protobox structure in the end user API so I mean the way I see as of now is that that we should not basically release at least we should not be releasing the genetic protocol photobomb structures either as a package uh for other languages, but for C plus plus. We should not be exposing the header files which we generate uh for protobuf and we do release the header files. As of now.

B

uh The reason why we release these header files is because the we do also install the recordable headers, and so there was a couple of issues which was a couple of years, which was raised in one of the PRS. We removed the recordable header, but in another PR I think moment has raised some some concerns that it should be used.

B

It should be available, probably for the exporters to be able to uh uh I mean inherit from it or some something that we wanted to use that, and then we brought it back, but inside the recordable headers we are also including the headers.

B

So probably that means that we have to eventually release the protocol videos also so yeah, that's something probably we can revisit uh if we really want to release reportable headers, if not I, think we can probably remove it and around with uh all the protocol. Windows, probably probably I, think the place. This issue is still in this question: um let's see where it goes, but I think personally, my opinion was that we should not be exposing um both.

E

B

Headers and also the photography videos, but yeah I mean let's, let's see where it takes.

B

um The other one is specify the metric reader configurable limits. I think this could be important. As of now we don't have. We don't really put any memory constraints in terms of number of measurements number of Matrix. We are going to generate number of Matrix, which we are going to store in our internal storage.

B

So there may be a problem that, over the course of time, if the collection interval is very long, we may our internal memory usage can keep on increasing, based on better based on the temporarily, whether it's, uh whether whether it's a delta or cumulative based on that I, think it's possible that our internal storage can keep on increasing.

B

If the polling interval is very high, you know, so we have to probably test those scenarios and it would be good if Patrick data can add some constraints in terms of definitely I mean how many time series data given instrumentation Library can only different time series that are given instrumentation that we can uh configure and for for a given instrument. How many time series data for a given instrument it can be configured default, is 2000, but I think it can be explosive over the course of time.

B

If you don't have such constraints, so that would be important to really uh have a very, irrespective of this PR I think we definitely have to see. How can we put some constants inside our it's called C plus plus configurations to see if you can um ensure that our memory does not keep Sonic exploring, cannot keeps on increasing and it can explode over the course of a few days. So that's something we have to definitely.

B

The curve one is make sure it is very clear that we are not building blogging API yeah this. This is I, think we are clear. I mean as of now that the currently logging spec the current logging API specs, is not for a logging API. It is for the appenders to use that API defined API to use through appenders um for the existing garbage liabilities of probably uploading the logs from open, Tech, Elementary logs to say logs for CG logs for cxx or any other uh a learning library.

B

If you want to do that, so probably for that purpose, I think this API is defined not not to be used by the instrumentation liabilities.

B

So you know right now, I think that's the discussion which which we are having in our current ER, don't know the latest PR, which was released by owent.

B

You can just quickly go through that yeah, this one so I think I from it. It looked like that we are. We are removing uh the existing logging API and we are just adding the new API as per the EPA specs, which is not the right to approach so yeah.

B

We have been discussing about whether we should use internal Lock, Storage or not so, and also another discussion is whether we should we should keep the existing API, because that is something which tells you might actually would be using in a new API would be only used for recordables, so yeah I think something moment is talking about that, but yeah, that's something. Probably we have to take care of in our implementation.

B

Another one is still in discussion, I think add Force flush in the tracing provider base. Api I, don't know whether we support it.

B

I'm just checking, for example,.

B

Okay, we don't do it obviously, uh so this request was.

B

To add it in that is API I. Don't think this should get added, but yeah I think that follow that follow the discretion, because I instrumentation library is not the right place to to plus the decrease it should be the SDK or the application, the application code, which should be responsible to decide when to do their real mesh, but I think probably let's, let's probably uh follow up on this. If something has to be done,.

D

B

Anything else, probably in the specs, otherwise we can talk on the existing issues. um Just let me know the information you can always talk afterwards. If there's anything to me either.

B

I haven't seen that or TLP with grpc built this is it Centos?

B

Okay, we haven't test, I mean I, haven't tested with sentos, but I think it should work. It should not break for one of the distribution and not work for any other distribution um yeah. Our code is not just dependent on any of these specific distributions unless until it's, the version of the compiler.

E

B

E

We already have a Docker file for Centos. Yes,.

B

Exactly and I think you did mention it somewhere in some of the comment. If you can use it, you know, but we did do we I mean we don't build that Docker file anywhere right as part of any CI.

E

No, no, we don't.

B

Yeah, so we don't know whether it's still building the latest Matrix or not, but yeah I think these guys they can try out that and see if it works, and they can try to use that instructions too. If it works, they can use version section to build it for the sentence.

B

Yeah, okay, probably I, think we have to see just.

E

To test it out, I.

B

Don't think it should it should fail, but probably good to test it for other distributions. Also.

B

Mm-Hmm, to enable ntls support on I'm, not sure what mpls is I know about TLS I, don't know how what exactly is empty in this, but yeah there is something which is supported by grpc, probably I, think it's good that we can also write a support of that I won't say it anything that we have to do it urgently, but.

B

C

Sounds like this is more like a config from our side right like if.

D

C

Is important for that, maybe we just add the complete flag or.

B

Something yeah we can do that. The way we do it for TLS I think we can do the simulator in this um yeah, so I mean I, don't know much about what mpls is so I cannot really comment, but yeah yeah.

C

C

B

E

I think the pr is.

B

There we have to just wait for them. Have we already discussed about insecure connections where we don't support it for HTTP and.

D

A

So what would you take to actually do it for HTTP? You have no idea how much could change. uh We are looking at.

B

um I think the change which I feel is basically, so we have our own wrapper on top of HTTP right. The HTTP client Library, which we use. So we have to add the hooks in that Library so that it's as part of the https handshake Whenever there is a request coming for the root CA certificate. That hook should be able to provide those certificates.

B

A

You know if this HTTP Library supports, as I said already.

D

B

It does not I mean support this; it will support SSL by default without without doing uh without validating the certificates. It will definitely support it. Just use https, it will support it, but when we, if we want to validate the certificates it won't it cannot, it won't support validating this. It will fail at that time because our wrapper HTTP client cannot provide those certificates for it to uh for it to validate it.

B

Okay, I mean that does that does it make sense, I mean I, I, don't know if for curl, if it can automatically read those certificate from the default location, just like curl, client, curl uh command line client. If it can do it, then it's well and good it probably. We don't need to do much.

B

uh No, but if still we cannot, it cannot work because we have to provide a certificate to environment variables, and then we have to ensure that our HTTP client Library should be able to load those certificates uh provided through command line and provide those to the coil client. So, okay, I will probably let me.

B

B

So probably we have to add I mean that's. That's probably my take. We have to add in another session State here uh for the SSL handshake. We have to trap, probably the the session State should be set whenever the request is coming for this certificate, whether it's client or server. Based on that we have to set the system State and then uh there should be a hook which should be called to upload to load that certificate.

D

So if you're doing something.

B

Like this all connecting we are providing, we are not using it anywhere, but I think, probably somewhere we have to uh in the event handler, uh which will provide a hook who supports it. But let me let me just see I think public.

B

D

Was a PR raised.

B

For this, I just want to see.

D

B

Me just search the full request. You think.

D

B

D

B

Was some work done probably disposed in a different way I? This was not complete, but I think the directions were good in the right direction to support it. Probably somebody can have a look into that.

B

Yeah, so probably I didn't receive something which would be a good startup point if okay.

D

So we are gone.

D

With the opening points.

B

It should be close right, I think I, remember! Oh, this is still open, honey.

B

I think this looks good. Let me know somebody want to review it. Otherwise I think probably I'll be at my end of today. I think I'll merge this. It's a very small change, just in case you'd be interested to see.

D

B

Pls setting for otlp yesterday here we already talk, work talk about this.

B

Your multiple instrument call back here which we have to support. It I think this part we can discuss in Department of Matrix our paid package for routines Elementary. There wasn't some comments coming for by Sergio. I'll just have a look into that and we had some issues in building it for RPM I.

D

Think I didn't.

B

Test it for RPM, so probably I was expecting something something not to work for RPM I just tested for him. um So we are. Let me see this. These are the good comments to test.

D

B

Yeah, this I think we'll wait for one more or we can do it in the next release right or once the release is done. We'll do.

A

It again, one storage is done. Yes,.

B

Does not support yeah I think we have to have this. It's not very urgent, but probably you have to pick it up. Sometimes.

B

So we have a PR for this invested waiting for this guy to sign the client.

B

Yeah this is there for Matrix firework block. I, think this is the problem. You have it so I'll just leave it to you see, maintain the shape point of instinct yeah. We, this is somehow we are.

D

Not able to implement that here.

B

And provide a mechanism for external development, so I think this is uh for Matrix. We are going to implement it in a right way.

B

Yeah I think this we have already discussed.

B

We can quickly see the Milestones I think probably.

B

For locks, there is a one. There is a PR rest as of now for blogger and blogger provider, changes and sorry logger provided a logger context. I need to see so yeah this one. No, this was this one, and even this one power changes.

B

Probably it is something I think I am reviewing it, as of now feel free to um check that we are I, think that that PA would be important to set a context of our initial set of apis, which we are going to use. So I think it's important here. You know nothing good. If we all can review it.

B

And yeah, so that's only one period as of now deadline is already passed by just just want to add some deadline. It should not look past due date. Probably 15th of I mean just just extending it to say. 15Th of December, uh probably I feel that it will take some time to review the changes.

B

Iteratively and all the pr should be one after one. So it will take some time so I'm.

D

B

It on for 15th of December now let me know I think if you will otherwise we will change it.

B

uh Matrix I think uh you know, lots of things are in place which we have to start working um I, don't see anything urgent as of now but I think slowly and slowly we should start taking. These I'll also have started.

A

The whole instrumentation that should be done.

B

Which? One in.

A

The middle the cleanup.

B

Is done right, it's not.

A

There, yes, but it's released.

D

D

E

This is done, yes,.

B

It was it right. Yes,.

A

Done in 170.

B

Division types add: support for exponential Matrix provide Builder to expose SDK, internals yeah, adding memory exporter, add a state pointer, obviousync instruments. We need supply dependencies for both getting started. Documentation, handle gaps, recordable interface between metrics storage, Explorer, basic education does not do Revolution support political instrument contacts.

B

You know, I think, apart from the last one, I think most of them at least we should be able to do it. This will need some design changes and I. Don't have much idea, but we.

D

Are good, as of now.

B

Because I mean none of the other languages support this as of now so I think we are not alone. Even though everybody is real, you know most of the languages being GA. They don't support it so fast that things should be okay for now, but yeah. Apart from that I think most of the things we know how to do it.

B

This may need some work, some effort, but I think eventually we can. We should be able to do it but yeah. It is probably let me start taking them one by one. I think lots of things are in my plate and I should start spending some time on this.

B

For the PRS, then you can discuss.

B

Okay, this is for C, plus plus 11 right, which is which I think we. You are working on this right now, right.

E

um This is done just the other. Pr has to be merged. First.

B

uh Okay, so this one is this: one is first, this looks good.

D

B

um So I mean only I got surprised that so that means that, with the latest upsell we are still able to build it with C plus 11.. Is that what you're saying.

E

Yeah yeah, it works.

B

Oh, and even though they, even though they with the latest episode, means the latest version coming from the uh from.

A

Here right, the latest episode is documented, as the last one for C plus plus 11. you're.

B

Talking about the last one, not not this, okay or the latest, yes,.

D

B

Fine, uh but the only I mean not a bigger issue. Only thing would be that our cmake build would be using a different version of Excel, and our bazel would be using a different version.

E

B

Because Civic I think we implicitly use whatever is coming from the third party sub module of grpc yeah.

E

B

Welcome, let's see I think it's not very important, I think it should be okay as long as I mean I'm. Okay I would probably prefer if we are using something which is coming ah inside the sub module, so I think it's fine I mean but but yeah it's your take. If you want, if you want to make it consistent in both feel free to raise a PR for CMA function,.

B

Okay yeah: this is.

E

um There was a memory leak, not erase I said last time.

B

Is it part of this PR? Yes,.

E

It's fixed here.

D

A

Was it the cleanup of photographer? Yes,.

B

Oh nice yeah thanks.

E

There is still one memory link for for the benchmark that.

F

E

That's coming from appsail and I couldn't find a way to fix it. I.

A

E

We haven't built the paint for Walgreens.

B

Oh, is it: where did you disable? It.

E

It's in like the first file in the CP in the CC file, and there is a benchmark well.

B

E

B

This one, okay: this is what you have done: no debug and enter yeah.

E

Yeah, that's it I've, just disabled the band for that.

D

B

Alex having.

B

A memory leak, foreign.

E

B

E

B

False, oh okay, not sure what what exactly yeah.

E

I'm not sure what it is.

B

Have thought it looks like I mean, so we.

F

Know yeah that.

B

Was that why we commented it out.

E

Yeah I I think good point. Yes,.

B

Yeah I think looks good I. Think probably just add that comment and I think that feel free to merge. It.

E

D

B

And this one is for eign.

E

And if you go to the workspace.

E

um It will get the open Telemetry from from main branch.

D

E

So it's really like the end user.

D

B

Okay, I'll go through that and probably it will take some time to worry, to understand how it is working but yeah. Let me go through this and then the.

E

Rifles C, plus plus 11 here.

B

Okay, thanks uh Hassan I think this looks good. Let me just review this.

E

B

Your tailway sampling, yeah, I, think probably Tom. This will come to you. uh You need not review it as of now I'm, just working with the team which needs that uh if they feel these changes are good enough, then probably I think I'll need your help in reviewing it. uh Okay, I mean, as of now, just just need not need not spend. Time on. This is.

C

B

Who lead ah there yeah if you're, not there I'll, add you in that discussion. Okay,.

C

Let me double check if.

B

You're, not there I'll, add you in that discussion. Okay, so probably you can follow it up. They said that they are going to test these changes. If it works, fine, then actually they will.

C

Probably be good too: okay, yeah.

B

But yeah we don't support the problem. Is we don't support tail based sampling, as of now in open television, C plus plus? That's.

C

B

uh So in our in open Elementary, C plus plus, the sampling decision is made when the span is started, but they may be the they may be requirements when we want to do something decision right before the span is exported based on the fields which have been populated after starting the span.

C

Support of the filtering in the ancient effort at the agent or.

B

Yes, we, uh oh yeah, we do support I, think the agent when you with the agent, you said which agent I mean.

C

I mean so for ucw. That should be like the more information.

B

Yeah, no I, don't think it's supported in monitoring agent, uh but it's it's definitely supported in open Telemetry collector. It has a corporate Elementary collector has exhaustive support of tail based sampling. Even it supports caching, caching, the spans from all the relay from the from the parent to the all the child spans and then deciding based on the. uh If any of these fancy scenario of any of this pan has this status as failed, then it allows you to to drop all the Sam all these fans coming from the same parent or something like this.

B

E

B

So you can cache it and you can make the sampling decision for all the spans which are cast based on the status of any of the any of this man.

D

B

That that support is anyway that cannot be supported in the SDK. So that's something which is part of collector, but.

E

In open television.

B

C plus plus I think we don't have a full Place support of tail based sampling. Obviously yeah, it's the it's because of the current design. We don't. We don't store all the attributes in the SDK. So all the attributes that directly serialize in in the exporter, so that that something which is necessary but yeah, that's okay,.

D

C

Asking about H and the preconception, H and I heard about that. The Asian and condom are like more advanced the filtering like it. Can it implements the query language which can do like the complex.

B

C

Condition apply that to other choices and then oh.

D

B

Yeah, that's that's good to know, I think that something which they can use it yeah so yeah, probably I'll. Add you there I think. If you want to add any comment. Do do add your uh comment.

C

Okay, yeah sure.

B

They may not I I think they may not be aware of that uh possibility. So yeah.

C

B

C

Good question: thanks.

B

Yes, we already waiting for this PR the CLA check. uh Let me see, if he's still doing it for things.

B

Okay, yeah all right: let's wait for some time that is already I. Think we discussed uh spin lock of uh yeah I mean feel free to just have a look into that.

B

You know this, we have discussed Legacy, we have already discussed. This should be closed right, yeah.

B

Which is exporter build? This is again also for CLA, it's kind of stuck. Let's wait for some more time, otherwise, I think we can add a picture. Any one of us can take these changes in order, because this I'll have a look into that there are. There are comments from ovent and also from new comment from Sergio, so welcome to this. This is something actually we have to implement so yeah I. Think that's! That's all we have in this and uh going back to this agenda.

B

Yeah I think this is important discussion plan date for next release.

A

D

A

I'm asking so I can we can finally remove the code after our deprecation.

B

I mean I think it's a good time to release it. We have already done the cleanup of the old Matrix implementation, so probably I think it's a good time. If one of us can take responsibility of doing a new, build I mean I mean that's somebody's interested just, let me at home or massage or anybody who's interested to have a new brain on. If you want me to do it, just let me know: okay.

E

I can do it.

B

D

C

Thank you, yeah thanks.

B

And the release so the last one was 1.7.0 right.

B

Should we do it, but uh I'm not sure I think we can keep it 1.7, but one also even 1.8.? Oh, if you think the duplicated uh old Matrix as the kind of baking changes, uh probably you can have it. 1.8.2 I mean if you take that as a breaking changes, which is fine, I, think, probably it's a major change. We are removing uh remove an existing functionality.

E

A

Yeah no no strong opinion both works for me: uh okay, um I I, don't know if we added more things in the metrics area since 170.

B

We did add, but I think these were the small non-breaking changes to cash. First, most of the curves related to fixing the crash I, don't think any change which should which should add, let us add the major version or something I mean most of the fixes were straightforward without breaking the SDK and the API interface.

B

So I'm just quickly looking into that.

B

But I don't see any major change. Yeah.

D

B

Let's say I think, let's plan for this week, uh I think any day is fine. We we don't need to wait or probably, if you want, there are a couple of PRS which I think, let's wait for if you want to at least the pr which was raised by oment. This is something which we can do it in this release. These will not go uh which are waiting for this for the CLA uh sign up. So at least we can really.

B

We can include this session log for this OTP OTL PHP client be part of this, and even this one is okay. You're about to the latest 1.48.1.

B

You can decide when, when it's the right time, you want to do it.

E

B

Yeah but let's let's target for this week and probably you need in basically.

A

Well, this week will be Thanksgiving, so if you're taking some time off, maybe a week.

B

After that, that's yeah, that's a good point, so yeah, but it's fine I, think probably probably next week should be okay and that's not already this week, I think it's okay for next people, I'll keep it to you. Whenever you want to decide.

E

D

B

There won't be much time actually at least Thursday and Friday I know. I know. Most of us will be out so.

B

Yeah and in continuation to that I think probably the storm. uh If you want to look into this.

B

I know some issues people have raised because of this, because this one is still uh building with.

B

With the preview of Matrix, so.

B

Okay, we're still doing.

C

F

C

F

C

Turn it off this.

B

D

Probably I think we can.

B

Do it now also if we want to wait for the new release, I think we can do it as you feel, but I think it would be good to remove it.

C

B

C

One we can, let me try and remove it as soon as possible. Here, nothing.

F

C

A new, very nice.

B

Yeah I I do see some issues raised in our discussion that they were trying to build with VC package, and then they were finding deprecated Matrix.

C

Okay, yeah and.

B

Then they were not able to print it. Yeah yeah.

C

D

C

B

Okay, anything else for this question.

F

E

I just have a question: you've got this HTTP notes and g-test yeah dependency.

E

I just can't recall what was the issue.

A

E

A

Was dependency on g-test and gmoc yeah.

E

Yeah, if we don't build with with tests on there, shouldn't be a problem right.

C

I think so yeah disabled has should should not really now those.

E

So the question is uh why this is an issue of if we are building with tests for tests.

E

It does have a dependency, the dependency and gmoc, but it does have it when only when you build with tests. Otherwise.

B

This is something which which has dependency on Gene Mock. Then probably we should move it under the test directory I mean this is not the right place for this to be there uh and we should not. We don't need to release it. Okay, we can release it for test, but we don't reduce it right in general if it is uh okay anyway, if you want to release it otherwise, but probably I think if it is something which is only for testing I think it's. This is not the right place for for this.

B

If it is just kind of a mock mock Library, maybe we can just probably put it in the room so in the test directory. So.

E

Okay, yeah, that's good.

B

That's yeah just add some confusion. You know it.

A

Could be very dependency, was in the in the cmake, also yeah.

E

It is, but it is there when you build with tests, not not for normal being yeah.

B

Yeah we are adding this check, which is fine, yeah but I. Think probably you can move it to the test directory and let it that is stay there. This is something belongs there. Yeah.

E

E

It should be there.

F

D

E

The only thing that I need to fix, then.

E

B

Okay, you know, if not anything else, probably I, think thanks for joining and uh let's meet next week, November 30th and yeah thanks everyone for joining every.

D

B

Can have a good Thanksgiving.

C

Happy holidays.

D

C

E

C

Thank you, bye.