Increase default value of net.ipv4.igmp_max_memberships
The linux kernel default for igmp_max_memberships is 20 [0]. For vxlan project networks, it is possible to specify a range of multicast addresses for use with VXLAN tunnel broadcast traffic. In the case where a range is specified a unique multicast group is assigned by neutron to each vxlan network, and once igmp_max_memberships is reached for a host, vxlan interfaces will silently fail to join the multicast group. This results in instances being unable to DHCP or ARP, and errors in the neutron agent log [1]. This patch increases the default value to 1024 which should be sufficient for all but the very largest deployments. [0] https://sysctl-explorer.net/net/ipv4/igmp_max_memberships/ [1] http://paste.openstack.org/show/796504 Change-Id: I24a0ca3001098b278d341f710ca2e4f00dc3e09a
This commit is contained in:
parent
7f26a27140
commit
7cbf51170c
|
@ -104,6 +104,7 @@ openstack_kernel_options:
|
|||
- { key: 'net.bridge.bridge-nf-call-ip6tables', value: 1 }
|
||||
- { key: 'net.bridge.bridge-nf-call-iptables', value: 1 }
|
||||
- { key: 'net.bridge.bridge-nf-call-arptables', value: 1 }
|
||||
- { key: 'net.ipv4.igmp_max_memberships', value: 1024 }
|
||||
- { key: 'net.ipv4.neigh.default.gc_thresh1', value: "{{ set_gc_val | int // 2 }}" }
|
||||
- { key: 'net.ipv4.neigh.default.gc_thresh2', value: "{{ set_gc_val | int }}" }
|
||||
- { key: 'net.ipv4.neigh.default.gc_thresh3', value: "{{ set_gc_val | int * 2 }}" }
|
||||
|
|
Loading…
Reference in New Issue