--
You received this message because you are subscribed to the Google Groups "kubernetes-sig-network" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-ne...@googlegroups.com.
To post to this group, send email to kubernetes-...@googlegroups.com.
Visit this group at https://groups.google.com/group/kubernetes-sig-network.
For more options, visit https://groups.google.com/d/optout.
To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-network+unsub...@googlegroups.com.
Either way, I would like to explore "LoadBalancer as it's own resource" idea
sooner rather than later because it would allow us to address existing problems with LBs that would otherwise require handling complicated cases in production clusters such as:
* better LB naming (https://github.com/kubernetes/kubernetes/issues/69293 & https://github.com/kubernetes/kubernetes/issues/29789)
* not excluding the control plane by default (https://github.com/kubernetes/kubernetes/issues/65618) - this one can actually be done today but I'm hesitant for reasons* topology aware load balancers / excluding nodes from LBs on a per-service basis (right now if we label a node, it gets excluded from all LBs which can be undesirable).
>> >> >>> To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-network+unsub...@googlegroups.com.
>> >> >>> To post to this group, send email to kubernetes-...@googlegroups.com.
>> >> >>> Visit this group at https://groups.google.com/group/kubernetes-sig-network.
>> >> >>> For more options, visit https://groups.google.com/d/optout.
>> >> >
>> >> > --
>> >> > You received this message because you are subscribed to the Google Groups "kubernetes-sig-network" group.
>> >> > To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-network+unsub...@googlegroups.com.
>> >> > To post to this group, send email to kubernetes-...@googlegroups.com.
>> >> > Visit this group at https://groups.google.com/group/kubernetes-sig-network.
>> >> > For more options, visit https://groups.google.com/d/optout.
>>
>> --
>> You received this message because you are subscribed to the Google Groups "kubernetes-sig-network" group.
>> To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-network+unsub...@googlegroups.com.
Hi SIG Network,I wanted to follow-up on yesterday's discussion around Service Node Exclusion. There were a few concerns raised about moving forward with this feature:
* Promoting Service Node Exclusion also implies that master nodes will be automatically added to LBs (this is desired). Should we also automatically include nodes that are marked unschedulable? I think the consensus was yes.
* If we do automatically allow masters, does that put master nodes at risk of self DoS in production because we are suddenly adding load from LB traffic there?
* How much of this logic do we want to standardize and how much should we delegate to providers? I think longer term we want to delegate more to the providers but working with what we have today we may still need to support something standardized.
* With v1.14 code freeze coming up, we won't have adequate time to test this change.
>> >> >>> To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-ne...@googlegroups.com.
>> >> >>> To post to this group, send email to kubernetes-...@googlegroups.com.
>> >> >>> Visit this group at https://groups.google.com/group/kubernetes-sig-network.
>> >> >>> For more options, visit https://groups.google.com/d/optout.
>> >> >
>> >> > --
>> >> > You received this message because you are subscribed to the Google Groups "kubernetes-sig-network" group.
>> >> > To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-ne...@googlegroups.com.
>> >> > To post to this group, send email to kubernetes-...@googlegroups.com.
>> >> > Visit this group at https://groups.google.com/group/kubernetes-sig-network.
>> >> > For more options, visit https://groups.google.com/d/optout.
>>
>> --
>> You received this message because you are subscribed to the Google Groups "kubernetes-sig-network" group.
>> To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-ne...@googlegroups.com.
>> To post to this group, send email to kubernetes-...@googlegroups.com.
>> Visit this group at https://groups.google.com/group/kubernetes-sig-network.
>> For more options, visit https://groups.google.com/d/optout.
>
> --
> Sandor Szücs | 418 I'm a teapot
>
--
You received this message because you are subscribed to the Google Groups "kubernetes-sig-network" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-ne...@googlegroups.com.
Isn't that filtering behavior for the master separate from the annotation?
See https://sourcegraph.com/github.com/kubernetes/kubernetes@master/-/blob/pkg/controller/service/service_controller.go#L588.
The code for exclusion includes role master, unschedulable and the annotation. We would have to change more than service exclusion -- we will also have to have introduce the label on existing master to avoid change of behavior for existing installs.
Yes, this can have unexpected consequences, for example, security issues due to exposing master to traffic.
If our final aim is to delegate to cloud providers, then wouldn't we not want to standardize this? As standardization will require a deprecation cycle to remove.
IMHO, this is definitely worth getting a KEP and holding until the 1.15, as it has potential for non-trivial semantic change (VMs added to LBs). It would be good to gather feedback on a KEP before implementing.
Even if we push this onto providers (which seems correct from a
who-owns-it perspective, but maybe obnoxious from a who-uses-it
perspective) the inclusion of masters is markedly different than the
exclusion of nodes. Concretely, we can't just suddenly announce
"masters are now included" without some transition. During that
transition, we need to make sure users get to choose. In other words,
we need a flag or something that says
"default-filter-nodes-for-service-lb", which defaults to true. We can
announce intention to change that default in 6 months, but in the
interim we have to respect existing configs.
I hate to make this hard, but we really have 2 questions in play, and
one is harder than the other. Maybe we should handle them as 2 KEPs.
>>>> >> >> >>> To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-network+unsub...@googlegroups.com.
>>>> >> >> >>> To post to this group, send email to kubernetes-...@googlegroups.com.
>>>> >> >> >>> Visit this group at https://groups.google.com/group/kubernetes-sig-network.
>>>> >> >> >>> For more options, visit https://groups.google.com/d/optout.
>>>> >> >> >
>>>> >> >> > --
>>>> >> >> > You received this message because you are subscribed to the Google Groups "kubernetes-sig-network" group.
>>>> >> >> > To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-network+unsub...@googlegroups.com.
>>>> >> >> > To post to this group, send email to kubernetes-...@googlegroups.com.
>>>> >> >> > Visit this group at https://groups.google.com/group/kubernetes-sig-network.
>>>> >> >> > For more options, visit https://groups.google.com/d/optout.
>>>> >>
>>>> >> --
>>>> >> You received this message because you are subscribed to the Google Groups "kubernetes-sig-network" group.
>>>> >> To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-network+unsub...@googlegroups.com.
>>>> >> To post to this group, send email to kubernetes-...@googlegroups.com.
>>>> >> Visit this group at https://groups.google.com/group/kubernetes-sig-network.
>>>> >> For more options, visit https://groups.google.com/d/optout.
>>>> >
>>>> > --
>>>> > Sandor Szücs | 418 I'm a teapot
>>>> >
>>>
>>> --
>>> You received this message because you are subscribed to the Google Groups "kubernetes-sig-network" group.
>>> To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-network+unsub...@googlegroups.com.
>>> To post to this group, send email to kubernetes-...@googlegroups.com.
>>> Visit this group at https://groups.google.com/group/kubernetes-sig-network.
>>> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups "kubernetes-sig-network" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-network+unsub...@googlegroups.com.