Antrea Community, 5 Jul 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Antrea Community Meeting 07/05/2021

Description

Antrea Community Meeting, July 5th 2021

A

Perfect so good morning or good afternoon, uh this is the andrea community meeting and uh today is uh tuesday uh july, the 6th uh or, if you're, joining the u.s it's july 5th. But since it's a public college in the us, there is no one joining us from there. So today is only july.

A

The 6th- and uh this is going to be- let's say, a shorter meeting, likely as we don't have any attendees from the us being a public holiday there, and also we don't have any topic which has been booked in advance for the agenda. Therefore, for today it's a pretty much open discussion and if you have anything to propose anything to discuss anything, that's bothering you please go.

A

Ahead, uh there is a silence, um I'm just curious uh if there's been any follow-up discussion to the multicast conversation we had a few weeks ago, um but I don't think that the people that presented the multicast area.

A

A

Is there any particular any bug that you would like to discuss.

A

B

A

Okay, thanks chan, uh I was just looking at this uh as an issue that uh jay shared over the uh select channel over the uh over over this lakshmi's luck channel on on sunday, and it seems that uh we have never validated kind of antria with the multiple nix, which seems a little bit strange.

A

It seems that if you have multiple nixon nodes, the actual gateway might not match the cube. The the ip address reported by the cubelet is that correct.

C

Yes, I think so I think currently, the only way to make a window traffic work is to guarantee that the node ip, the interface that has in the node ip, is the same interface that has a default route. Yeah yeah.

C

I had a brief discussion with wayne this morning a few hours ago, and we are considering that uh maybe it's possible to decouple this logic, that and light like that user to specify a interface that could be used for data paths. So oh external traffic and all encapsulation traffic can go that uh data link and that detailing data link can be different from the uh device that has the node ip yeah. You guys think this is.

C

This makes sense.

A

You know it makes sense to me, but.

A

Micros applies to windows only or does it apply to linux.

C

As well- yes, that's a good question. We also discussed this and think this applies to linux as well, because even for now, when I wanted to test linux performance, I had to make the the the device that has uh higher uh bandwidth uh to hold the node ip, so it's very inconvenient to performance to switching the node ip.

C

That's right, yeah, because I had two interfaces. One is the default and it can be reachable from my laptop, but it has a lower bandwidth and another one is right: network private network.

C

It has higher bandwidth, but since it's not, it cannot be reachable by default. It's it makes me how to make some uh configuration to to use that that that ip for kubernetes, controlling yeah.

A

Understood: okay, so thanks for uh for the clarification here.

A

There is, uh there was another thing that uh it's uh uh that I wanted to bring up today and it's uh kobe's work on uh the no different, integrated node diaphragm controller um copy. What's the current status of your pr.

B

I think I've addressed all the uh comments which uh we had on that pr and uh I was I was doing uh some testing on it like a week ago, which, as far as I recall, because since then I had uh an openstack break, uh seems to look okay.

A

Okay, but you uh are you testing it with the like a standard, kubernetes environment, or are you able to test it with an open shift environment.

B

I didn't test it with double shift.

A

Okay, that's uh that's good.

B

A

B

Maybe it is worthwhile to to give it a try with.

A

uh Yeah, I mean in theory nothing should change, because you know uh the controller. The controller should behave in the same way is just that. uh The reason why that pr has become important is that we have been painted on the community channel by some community members that are trying to install.

A

I don't know if open shifter, okay with andrea and uh you know, since openshift version 4.6 um the they have disabled the node ibm controller and in cube controller manager, which means that uh the only way for making andrea work is to go to openshift before it deploys the cluster and tweak the bootstrap node to re-enable the node iphone controller, and then andrea deployment can proceed. Obviously this is less than ideal, because openshift deployment is supposed to be fully automated and instead you need a you need to go and do this arc.

A

So that's why that pr is a kind of important at the moment. Okay, so it's good that all the all the comments have been uh have been addressed and uh and yeah, and then eventually we need to have a, I think, eventually kind of quite soon we will probably need to have a conversation. Also about adding uh uh openshift uh to our ci upstream ci pipeline switch.

A

To be honest with you, I don't know how that uh I mean, I know how that work. Technically, I don't. I don't know if we need to deploy okd or if we need to get some. uh You know some uh some account from reddit for for deploying openshift, because clearly it's it's not free and open.

B

B

A

Ncp no yeah.

B

A

uh No, no, we just need to run openshift within three yeah. No.

B

I said that if we do that with ncp, maybe we can do the same thing, but well.

A

But maybe ncp, you know, ncp is a completely different model because it's like a it's a corporate product. We are just talking about purely open source stuff. If we put something upstream, we need to have a license and uh and and and we should not use the ones that are used for corporate purposes anyway.

A

So we will try and uh and address this one as well and uh perfect. So, just out of curiosity, has anyone ever tried to deploy okd with andrea.

A

I mean probably not probably not- I don't know I yeah, I don't think so anyway. So that's uh that's all about this pr, so you know if, uh if you have some time, please have a look, also hobbies pr, so that we can uh try and get it in the next century release, and I believe that uh that is all that I had on my agenda for today is there anything else that you would like to discuss.

A

Waiting 15.

A

A

Perfect, it seems that there are no additional topics so as anticipated, this one is going to be a shorter meeting. So I would like to thank everyone for joining us today and, uh as usual, thanks for all your contributions and uh say that I think this meeting is over, so I'm going to stop the recording thanks again for joining and talk to you again in tweaks time.