You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Dec 5, 2017. It is now read-only.
kube-proxy accepts a port range from which proxied ports are allocated from. we need to enable this flag via scheduler config and in k8s/dcos packaging
The text was updated successfully, but these errors were encountered:
clarification: the kube proxy port range is not for node ports, but for ports allocated by the userspace proxy impl of kube-proxy. the stock k8s implementation uses port 0 to allocate userspace proxy ports from the ephemeral port pool.
in cases where users want to control the port range that the userspace proxy allocates from (read: they don't use the ephemeral pool) then they can provide a custom port range for this. allocating such a range via mesos offers (assuming NO ip-per-container support) is error prone at best: kube-proxy becomes launchable on a slave IFF the entire userspace proxy port range is available.
i'm lowering the priority of this for the following reasons:
kube-proxy may be run in iptables mode, which doesn't need to allocate any additional "proxy" ports
the default implementation of kube-proxy's userspace mode allocates ports from the ephemeral pool
mesos user's should have already configured their mesos slaves with resource port ranges that avoid overlap with the host/OS's ephemeral port range
kube-proxy accepts a port range from which proxied ports are allocated from. we need to enable this flag via scheduler config and in k8s/dcos packaging
The text was updated successfully, but these errors were encountered: