Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Use IPv6 multicast with IPv4 anycast? #465

Closed
Harvie opened this issue May 7, 2019 · 5 comments
Closed

Use IPv6 multicast with IPv4 anycast? #465

Harvie opened this issue May 7, 2019 · 5 comments

Comments

@Harvie
Copy link

Harvie commented May 7, 2019

I have cluster with IPv4 unicast adresses, but i've found my network can only handle IPv6 multicast traffic... Is it possible to use IPv6 multicast with IPv4 unicast setup?

Problem is that my networking hardware don't have IGMP querier (IPv4), but it has MLD querier (IPv6). Unfortunately IGMP snooping can't be disabled... But i haven't deployed IPv6 anycast yet. So it would make sense to use IPv6 for multicast only, until i get full IPv6 support in my network.

@jfriesse
Copy link
Member

jfriesse commented May 7, 2019

@Harvie No it is not possible. I would also recommend to reconsider usage of multicast because it is known to have a problems (mostly switches related) and udpu/knet has comparable or better speed.

@ReyRen
Copy link

ReyRen commented May 7, 2019

@Harvie In the corosync 3, the nodelist is required whatever use the udp,udpu or Kent. So multicast might be obsoleted.

@Harvie
Copy link
Author

Harvie commented May 7, 2019

@jfriesse @ReyRen really? I am using corosync which is part of proxmox and in proxmox docs they strongly reccomend to use multicast. Well they use 2.4.4-dirty version... I was thinking that the muticast is used in order to drop packet count for large clusters, because when N nodes "ping" each other, it's almost N^2 packets every millisecond or so... But i don't really know how this works internally...

@Harvie Harvie closed this as completed May 7, 2019
@jfriesse
Copy link
Member

jfriesse commented May 7, 2019

@Harvie Really. In corosync 1.x udp was recommended because udpu was really not ready yet. With 2.x we are recommending udpu, because of huge amount of support requests mostly related to bad/incorrectly configured switches. And with 3.x we recommend knet (which is unicast or sctp) and made nodelist required.

Also you are technically right, corosync 3 has no rrp. But corosync 3 with knet can use up-to 8 links in rrp like fashion (so it's actually much better). I'm also not that sure proxmox guys stick with 2.x for long time, you can ask @Fabian-Gruenbichler.

You are right that multicast should be more effective and it's true from perspective of sending node to switch. But receiving packets is same story no matter if unicast or multicast is used.

@Harvie
Copy link
Author

Harvie commented May 7, 2019

But corosync 3 with knet can use up-to 8 links in rrp like fashion (so it's actually much better).

Yes. I've realized this and therefore deleted my comment. Thanks for clarification...

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants