Open Telemetry Uncategorized, 16 Nov 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: 2022-11-16 meeting

Description

Open Telemetry Meeting 1's Personal Meeting Room

A

B

C

Say hi everyone.

A

D

C

I appreciate the screen.

C

C

Exchanges I haven't Consolidated today, uh but I do know that there are some changes which are still in discussion and there are some PR rates for this. For the specification for uh uh I mean for changes in the locks API. Basically uh splitting I mean or decoupling the event API from the logs API. That way uh need some changes in the PS, which is event is Raising, uh so probably I think uh that's something you probably need to see.

C

You just need to see how how in which direction this is going. So basically it's there should be a separate API for creating an events and for the logs.

E

C

Of now we have a single API to know that so, but that's that's something appear has been based and I just need to probably see you know what's happening in them and apart from that.

C

You know I'm just looking into. There are some other old specs changes which we need to look into.

C

You know I think we haven't seen Matrix, we have Matrix 20 convention, not specified I, don't think it's there were some discussions. I remember but I think it's not still.

C

Okay, I'm just looking into view of Destiny updating, okay, this is still closed.

C

um Okay, this is open deviation.

C

Yeah I mean in case just just please, please I think. Let me know if there is any specs changes which probably we need to discuss. I will look into that again, probably today, and maybe if something has to be discussed, I'll put put it in for the next meeting.

C

But yeah, but otherwise, let me know if something has to be discussed in there is something which you see would be a good tool to make it.

C

You can start with the pr. If this.

C

A half meter context get meter return a copy instead of fan here. This has got created today. I knew about this issue. uh Basically, we get D meters. So whenever we have to do the collection, we get all the meters from the meter context and then for each of those meters, we'll call a collect API, but there could be a scenario where the meters are still getting added, and uh so, even though we get all the meters- and this basically is a vector of restore internally as a vector, so we just take an awesome.

C

We just take a noise to the span construct on the vectors memory object and if we try to iterate it and it's something has got added that may change internally. Vector May relocate these memories, so it may change the behavior iterative that causes this segmentation point.

C

uh This PR is changing it to use a cop to do a copy every time we call get meters and then use that copy which will work. Fine, I, think that would be a sleeper approach, but probably I just wanted to see.

C

If we really should change and tell the stories of meters from Vector to something called linked list or something so that, even though, if anything, any new meter get added that we can get added at the end of the list and we can still iterate uh all the initial set of meters that should not cause any corruption. We already do that to handle pan processors. So I just want to see if we can use similar thing for here uh yeah.

C

But the problem only again comes for meters, because in traces and logs we never have a scenario where we have to call a. uh We never have a scenario where we have to get lists of all the tracers or all the loggers and then need to call same API. On top of that, this is a case where we need to do it for meters, so I think for this I have to figure out if you can change something in but yeah, just just very early in the practice. One probably I'll look into this.

D

Well, let's do it.

C

Pick spin lock, nuts for digital compilers.

C

Yeah, that's I think should be good if it has been already tested.

C

When I saw there were some issues with the formatting ornaments and using the incorrect version of the latest version of C9 format.

B

D

B

Words afterwards.

C

A new log record and reportable implementation.

C

Yes, okay, it's turning.

C

Okay, let me see his latest comment: I haven't gone through this, but I mean I just had an issue. The way we are creating read, write log record where we are going to store all the attributes inside the SDK and then from there we are going to serialize it, which is not something which we were doing earlier and even for traces. Also, we don't do it. uh So my comment was basically if we can deviate from specs by not having read, write, longer and directly use reportable to serialize the data.

C

But let me see I think we probably won't have some comments on that.

C

Once we decide on how we should be going ahead, whether we really need we, they will record and redirect login for them. Probably um this is a right approach. I mean only thing is that we never uh whether we I mean we can create readable log record, but we do. We really need get methods on the log record, I mean once we set the log report. Why do we need to use Getters again, I mean? What's what's the use case?

C

We do have that use case for Tracer, where we may want to read these pants. The span attributes which we have already written basically for the tail based sampling I, mean if you want to take a decision of sampling when just before ending the span that time we may do need to get some getter methods and try to see uh Summit. They take a decision on the sampling decision based on both attributes of a span, so those get emitted makes sense there, but for logger we don't need for logs.

C

We don't have any sampling scenario also, so why do we need readable log report? I mean which provides an API to get the attributes which have been deleted. Also, yes, but yeah. Let's discuss it here.

C

Yes, multi conversion upgrade to this should be good to go.

B

uh Yes, I think it should be okay,.

C

Okay, Tom, you had some comment right. I think this should fix. That I mean where we were doing delete.

D

C

Error, so we so this is I, mean.

E

The variable links have been changed.

C

Right now with something like this thermal paste, so I think this is just see that it's good. It looks good and if you want to approve it and then we can probably.

C

C

Excision Lockport hotel, please this should be okay, I think probably I think I had a view of this and it looks good I just had some minor comments. uh Probably it was possible that we don't need to use these variables uh the session variable they are same but I think it's okay, it's a very minor comment, always wants to maintain them separately. I think that should be okay.

C

Okay, this we have to keep it after a couple of a couple of releases, probably next release name, we have to merge. Yes, keep.

B

The pr open and merge.

D

B

Only not to in this release, but the one after about um I'm, not totally happy about keeping your appear open for a long time, because if we do that with many things, we we may have emerged uh collisions that we don't even detect until the last minute, but so far this is the best I could I could find.

C

Yeah I think that's this is this I mean I? Don't think this is a very general case, a very specific piece. Let me want to duplicate up to. After probably only one will do you see, things should be opening I, don't think there would be measures. Conflicts coming because of this conflicts, so should be fine for now, but yeah in general. I think we should not keep put near open up around.

C

Legacy grpc into entrance okay son I need to go through this probably I. Think we've updated that CI for this right. There.

A

Is one shoe there is um a data race that I need to find out why it's happening in in one of these Benchmark tests.

A

Let's see yeah, so it's in the last test, where we do some end-to-end test.

A

Should be the last one.

C

Not this, this.

A

C

This one okay, yeah.

C

A

And I also try to upgrade to the latest grp scene. There is a compile issue there.

D

A

So I wasn't able to compile it. One 150 once 150 that one.

C

But earlier we were trying with.

A

C

Okay and then it was compiling in the news: okay,.

C

And that compilation is used with the grpc or it is with.

C

A

Stuff that are not defined.

C

A

I'm not quite sure why what's happening.

C

Okay, but but with the 149, correct, 148 or 149., so.

A

Do we are currently using 148 of.

E

A

We make and this PR will use 148 for basil as well.

C

Okay, and with that, we are getting showing this yeah.

A

About 148, we have the race here.

C

C

And with the earlier version, we were not getting any base condition in.

D

C

No, this is strange.

A

uh One more thing: I had to um also upgrade the app sale version.

A

If you go to the no it's repository.pazal.

C

This one right, yeah.

A

So we need this for compilation also.

C

C

The new grpc version is basically dependent on this particular action version. Yeah.

A

A

Pc separately so.

C

Yeah, but what Excel version is the version which we use, which comes as a sub module with the grpc? Oh.

A

I I think so: yeah I.

D

C

A

But I think so.

D

C

Okay, so right now, this is not in a state where we can no.

A

C

D

A

Need to fix I mean I need to find the reason for the race.

C

Conditions, okay, probably I'll, also try it out. The latest version I haven't tried for the Basel, because sometime, let me try with the meters by maybe, but you see something there.

A

um One more thing just somehow related to this: there is a new release for promptius and they also stopped supporting C plus plus 11.

C

A

So if we upgrade the prompt use version, we have to do similar thing for C, plus plus 11.

C

That would be tricky, I mean similar the way we did for grpc. We have to do that also.

A

Maybe it's trying to. We also stop supporting Secrets yeah.

C

I think that probably we have to take a decision.

C

I think it definitely I think we probably we have some more statistics in terms of the the uh most of the I mean Linux distributions are still by default, or I mean still supportable, with the C plus plus 11 uh I mean somehow, if you can get the list of all these stills of distribution like Ubuntu, all the position which we are still I mean supported by Ubuntu assembly by debut, and then we see which all as children's, if we don't have anything, still supported distribution with C plus we'll have nothing.

C

We can always take that call, uh but I think good to get some data before really taking that call. That is really good. If you can get that or C Business.

E

C

Going with the trend, grpc Prometheus I mean if they are stock, supporting C, plus 11 I. Think that also is a one flag for us that we should definitely started thinking about that.

D

D

A

It's 40 uh 14.

C

81 today, another build issue is coming with the latest UFC right I mean.

A

So I'm here I was trying to compile with 48 148. 1, okay.

C

C

uh 1.50 but 1.48 build is successful. Right 1.48. It.

A

Was various condition which was yeah except the race condition.

C

And but we were heading a build issue with 1.15. Yes,.

A

C

A

C

One in this condition.

A

And it's 48.1.

C

We are probably let's discuss that uh if you can collect some more data around C plus 11., then I think probably discuss it next week. If we should decide, stop supporting this weapons.

C

And then, if we say I mean we agree, then probably we have to document, and you know change our series accordingly and yeah, but I think to have some data before doing it like I can I can take a call and I can at least I can collect some data across the different Linux distributions and property, also Mac uh Mac OS also and in Windows I. Think we already have data I've been probably shared that.

C

How do I close this.

A

C

For which is exportable, okay, I think this was good to go I. Think probably we could not. Oh.

C

Sign the CLA, if he's not doing it, probably we can ask him if it's he's. Okay, one of us can raise up here. It is changing. I mean we have done in done that in the past there are issue and it's a valid mix.

E

C

I remember: there was one comment: uh there was a issue raised if we can probably for the same for creating the debt packages, and somebody has commented I think if we can support it.

C

I'll look into that issue, but I think it looks like that there is a there is an ass coming and people may be interested. If you can start supporting this and or if you can do that, we again, we are not uh saying that we are going to deliver running or we are going to release any binaries. We are just facilitating, facilitating uh creating the Devon RPM patterns, so I mean please please both go through this I will feel it's good. We can definitely merge it.

C

Let me see if there was some comments please further from this I know. You know. Movement has some comments. I will try to resolve them and please review it afterwards, something if there is all the comments.

D

D

C

Okay, anything else in this, uh otherwise you can quickly see if the issues of carrying memory, exporter I think nothing much. We have to rewrite this.

D

C

I'll Define difference too I'm just looking into the issues and to find difference to this I think there is a PR based body within this track at further foreign.

C

There I think he's probably asking about otlp grpc exporter, because we don't support, secure connection for OTL P HTTP export on that and I think this is a valid concern. If it's the grpc and it's we already have couple of issues for HTTP exporter, we don't support, support, secure connection, we don't support uploading certificates and all the stuff, but for grpc I think this. This makes sense as per the specs by default.

C

It should be the secure connection, so we uh have to ensure that we try to connect securely uh and unless, until it uh we don't have those configuration parameters for secure connection, we should fall back to insecure connection or or error error out.

C

Insecure connection, only if so, it's explicitly specified in the options soon and.

B

Yeah, we should also change the environment variable that we use, because the spec has changed and now uh name this insecure. It was a different name.

E

C

C

We should be good, I'm, not sure which so.

B

We need to align on the environment, variable names.

C

D

B

Necessarily enable things like that.

C

It's not in this right, but.

B

We don't well here we have the old one we have SSL enabled and instead.

C

It should be okay or whatever. Okay.

C

D

C

I'll just mention it.

D

C

C

Does not really support it, so now you have to probably I'm just mentioning it again.

C

But we already have that as a help wanted in this CB should be good.

C

C

Support multiple instrument called back yeah I, just deleted. That I mean this is uh not a mandatory requirement from the specs, it says, may support, uh but I think would have an issue to track it and it won't be a major change, bigger change, I think um so I think probably and see if we can include it in one of the upcoming religions.

C

Yeah RPM package I just service talking about that's. Why not that package with RPM package for open television, so I think he commented the same way using CPAC the same way we are using. It.

D

C

Yeah I also noticed that generated purple packages contained the HTTP client was sent. Yeah we Drive does not fallen on gases, does not follow the name and also this library is dependent on g-test and gmoc and I. Think as on you know, you know this better. You could wrote this code, um so I think we have an issue for this I. Remember creating a issue on that uh that we have to remove this Gbox gtest dependency from nothing.

C

Some options we had, but then it's probably.

B

C

Is this dependency.

B

There I mean what is: uh is it used for.

D

A

Can't go away I added that.

C

I think it uses in our test. Cases- probably is you know the unit test and that five first dependency was added.

C

The history of this huh you're, probably from walking I, think for creating using gmog we're using gmoc inside this Library inside inside no STD I mean loss and HTTP sent and property for gmark. We are using it.

D

C

But yeah I did have some some idea on. How can we remove that?

C

um We probably will look into that.

A

I'll take into this I'll: do it yeah soon sooner yeah.

C

Yeah yeah whenever I think it's not very urgent. Probably once you fix even you have done your that basal grpc upgrade uh up to you whenever you want to see.

D

C

Okay and yeah this we already discussed, bring your own dependency. Let him walk, probably you will. You will.

B

Always follow us still estimate to test with the c make perfect.

C

B

Path and if it works, I would.

D

B

The issue, then so I still need to look at it.

C

What an open, Telemetry collector Zipkin receiver does not work with C plus plus, yes, I, think I looked into the specs and I found. It is a discrepancy we're not following this text. We have to send the text as okay, but we don't do it probably you're sending us a number one and zero.

C

uh So that thing we need to fix the server you can exporter. uh The Zipkin. uh Probably server still accepts that, so we never get any error using it.

C

But I think this has to be eventually fixed, because at collector level they don't support it as zero, and one said we have to ensure that we are consistent with the connector.

C

C

Okay, this is already PR. Is there yeah, so.

C

Where this I have a relationship issue, we think aggregation does not really application property. We don't know how we can support it. It's would be a bit complicated thing because we remove all the dimensions before and before doing aggregation. So when we reach the aggregation logic, we don't have any Dimensions, so we won't be easy to handle that, but I think it would be same issue with all the interpretations. They do the same way they remove all the attributes which are to be filtered out.

C

So at the time of aggregation they don't know which one attributes were there. Also the decision to re-aggregate based on those that.

D

Those removed attributes would be difficult.

C

I will walk, alert, yeah, I. Think Tom. Do you have this with you? Whenever you have time, please see um SEC fall due to corrupted instrument, scope, uh I, think there's already a PR, but probably I have to see if there is a better approach to under this.

C

Maintain the shared contractors is fine. I think we had some challenges right, Assam to support it. So I think it's.

B

D

About this very.

C

Subject you know stupid oh yeah provide a mechanism for external developer to distinguish between internal and public and clearing.

C

Okay, yeah I think you're really just supposed to have a stack metrics and as of now we'll ensure that we're not going to I mean it's not going to change exposure because created is East.

C

I think there's no logarithms, and probably we will like we're going to work on that uh and I think it has told you are very old one which we have to do clean up and have to do it. Sometimes we personally not getting time to book enough about once, which may not be still valid, but let me know if something I should be discussed out of this.

C

And yes, yeah: let's go back here.

C

In the beginning to clarified in contributing yes, I think we have, we should do it. That's a good point. I have seen this conclusion coming on the people coming when you are also.

B

So yeah so today someone uh filed the pr to fix the Intel compiler and.

D

B

There was a format issue, so he tried to apply selling format and then everything uh all the formatting went ballistic because it was very popular. It was a different version of the same format so especially for for someone who is doing a new PR, a new contributor I think we should make it very easy to contribute so that it decrease, increases the chances of more people in the open, telemetry, yeah I. Think that's that's a.

C

Good point: yeah.

C

B

I looked at it quickly, but we we don't say anything inside the script itself. We we pick selling ceiling format 10 if it is there, but otherwise repeating we pick slang format whatever it.

C

Is so I think it's coming internally from here? It's about this somewhere we can specify. We are doing it directly using the C language format, which which program they should use.

C

Your marks would like to create an issue for this and.

E

C

Track it down and.

B

C

Can see if, once whatever time can work on time, yeah.

B

It's just uh some update.

C

Mac OS compilers, uh our CIA.

B

Is disappearing, this just happened today and for whatever reason, if you look at the new PR, uh the the, where is the build break on Mac OS for on something new, it's not the format. It's my immigration well,.

C

It's not I'm. Sorry, yes,.

C

Good morning started coming suddenly now. Does it mean that GitHub has changed the Mac OS latest to some other version, I.

B

Don't know I, don't know if it's the compiler that changed or the compiling options or if it's Absol. If you look for error column, you will see it in the log.

C

C

Yeah I mean it can be a let's put it if it is outside change, then that should not, because we always take a given version of jio I've, seen that it should not change unless until we are always taking the trunk for building, we don't do that.

C

It should not be that I mean I presume it should not either it could be the compiler change. That seems like I, mean I, don't know if this I'm guessing.

B

Yeah look at the line 74.

C

C

We should change our I'm, not sure this episode is internally getting built from grpc right. We don't really build it explicitly it's creating wheel from here. This is wrong.

C

Okay, let me just see if it is.

B

So I've seen some recent changes in the runners on Mac OS, uh but it's from uh two weeks ago. It's not just today changes.

C

But we use Mac OS latest right latest means Mac OS, 11.

B

Well, the bead log says 12.

C

B

Well, with the the CI we've been failure.

C

um I'm gonna look.

B

At the operating system.

C

D

C

Latest and if it is latest.

C

For you from your, why it is.

C

But probably it's this documentation is not that's possible. I think the documentation is not updated in the readme and they have changed this latest two point: I mean this I mean when I'm trying to explain this documentation is not yet updated and actually Mac OS. They have changed their 2.2 matters, the leadership pointing to macro and square. Now it.

B

C

That's possible and because of that.

B

So this just started today so just be aware of it. If you see that in appear.

C

I can't see anything close. It is very recently probably first closed earlier and the release was done.

C

You know this is something different.

C

What could be causing it and if the words you can as of now, don't create it to Mac, OS 11 explicitly and then public debug it okay.

D

C

Fixing, if it's something related to that nothing so either we can keep using Mac, OS 11 or probably we have to see how can how can this? uh What I think is if it's an issue with the latest version of Mac OS? If the Excel library is not building, then probably we should see an issue at the absence GitHub also so I think we can track it there. Also with some. There is some issue based.

C

And let's see if we really want to.

C

Do something this probably will be fixed, discuss it over the issue or how to really find them. This.

B

If it, if it can be repeated first of all uh in case, we are changing what they just point to right now,.

D

B

Then we can, we can decide after.

C

Yeah, we can discuss it over there. Okay, probably you.

C

Okay, anything else.

C

Your Tom, your son,.

A

And nothing for myself.

C

Okay, if nothing.

D

C

Think we are good uh with the meeting today meeting let's meet next week. Okay,.

B

So, just if you, if, if you could uh review the change for cementing conventions so that it can be merged.

C

C

It won't be a bigger review, I mean it's just changing the variable instrument, s.

C

Are in Tom is mute or probably is some balloons I'll.

D

C

Up with Tom I think, if you can do it, otherwise I think we should be good to go to even merge it. I mean it is not having time information. Okay,.

A

E

A

Sorry there was a Mac OS version change.

C

Who is it okay,.

A

It's my in my PR, it's eleven seven.

C

Eleven seven.

D

C

A

um Even the open one like.

C

Okay, I think you can. Probably this is easy to see the just seeing what somebody did that and it does not show any checks here as well. I think it'll probably be better to go. There.

C

Which one you are seeing.

A

C

Level, 700, yeah, so I think it's changed. That's causing a problem, so I think there should be. Let's follow up. There should be a GitHub issue coming for, but.

E

C

If we get a issue coming for Excel because Epson they are just deprecating something right that should not be a determination, but then it's just the build is breaking because it's a some functionality has got it right. So probably, if that's the case, we may have to change our build to not error out on the application on replicative.

C

Okay, thanks, that is, that was helpful. Netflix probably continue discussing it so I.

D

Think we can raise up.

C

uh Issue I think it will.

D

C

Consistent in that case, I suspect the failure would be consistent. So probably if it is constant, let's raise a issue and let's discuss it over there, how to fix it, because that will block uh all over merge and I think even symmetic convention merge. We cannot do it without that, unless until we override it so.

B

Yeah I'm, assuming for.

C

Example, this is still okay, oh okay, fine then I think, let's, let's ensure that we don't do. We don't do any more changes in this. We are.

A

Change the Mac OS version in the CI.

C

Yeah, that's something we can all do that I mean unless and if good.

E

C

Always run on latest but I think we don't see any other solution, probably I think we can fall back to the earlier version. What Mac OS.

C

E

But I think, let's.

C

Discuss it over.

C

Okay thanks everyone for joining have a good rest of the day and the week and yeah. Let me meet next.

E

C

Next week, I just want you to see. uh We have vacations to next week is 21st right. Okay, Tom, do you know when okay we'll have we have a vacation on 24, 30 minutes.

A

C

Yeah, okay, then we are fine. Oh.

B

Is it what uh Thanksgiving already yeah.

C

We will be good for next week.

C

Okay thanks everyone for joining.

D

Thanks everyone thank.

E

You everyone bye.