Full HA(Ceph) keepalive error on loadbalancer

Bug #1295747 reported by Shweta P
18
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Cisco Openstack
Low
Unassigned

Bug Description

While deploying Full HA (without swift) the parameter swift_storage_interface is by default set to eth0.222 in the user.common.yaml which is not changed, since the deployment uses ceph and not swift.

But due to this the /etc/keepalived/keepalived.conf in the loadbalancer has the following entry
vrrp_instance 51 {
  virtual_router_id 51

  # Advert interval
  advert_int 1

  # for electing MASTER, highest priority wins.
  priority 101
  state MASTER

  interface eth1

  virtual_ipaddress {
      192.168.242.41 dev eth0.222
  }

The loadbalancer does not have the interface eth0.222 created by default and does not need it for a full ha ceph deployment. So it reports an error
ar 19 22:32:40 load-balancer01 Keepalived_vrrp: VRRP_Instance(51) Unknown interface !
Mar 19 22:32:40 load-balancer01 Keepalived: VRRP child process(6388) died: Respawning
Mar 19 22:32:40 load-balancer01 Keepalived: Starting VRRP child process, pid=6395
Mar 19 22:32:40 load-balancer01 Keepalived_vrrp: Registering Kernel netlink reflector
Mar 19 22:32:40 load-balancer01 Keepalived_vrrp: Registering Kernel netlink command channel
Mar 19 22:32:40 load-balancer01 Keepalived_vrrp: Registering gratutious ARP shared channel
Mar 19 22:32:40 load-balancer01 Keepalived_vrrp: Initializing ipvs 2.6
Mar 19 22:32:40 load-balancer01 Keepalived_vrrp: Opening file '/etc/keepalived/keepalived.conf'.
Mar 19 22:32:40 load-balancer01 Keepalived_vrrp: No such interface, eth0.222

Even the controller vip is not associated with any interface.

Workaround change the swift_storage_interface in user.common.yaml to an interface that is present in the loadbalancer(eg eth1)

Changed in openstack-cisco:
milestone: h.3 → i.0
Revision history for this message
Nagaraju-Bingi (nagaraju-bingi) wrote :

Change the swift_storage_interface parameter value from eth0.222 to eth0 in user.common.yaml.

Revision history for this message
Chris Ricker (chris-ricker) wrote :

changing swift_storage_interface to a present interface is a workaround for now

right fix is not to enable the swift proxy when not using swift

Changed in openstack-cisco:
importance: Undecided → Low
Revision history for this message
Chris Ricker (chris-ricker) wrote :

priority set to low for now since configuration to avoid is pretty straightforward

Changed in openstack-cisco:
milestone: i.0 → i.1
Changed in openstack-cisco:
milestone: i.1 → none
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Duplicates of this bug

Other bug subscribers