Re: [Keepalived-devel] multicast packets not sent by lvs director
Status: Beta
Brought to you by:
acassen
|
From: Nicolas <ara...@ya...> - 2005-04-27 17:01:42
|
Francois JEANMOUGIN wrote:
>Nicolas :
>
>
>>I have the same problem with keepalived 1.1.11
>>Any ideas ?
>>
>>
>
>It looks like your dameons try to get MASTER on both instances at the time,
>as if you included those in the same vrrp_sync_group. Try to define two
>vrrp_sync_group, each one with only one instance to force it.
>
>
>
Thanks for your reply Francois :)
So, i have appended this in the configuration of the two LVS directors :
vrrp_sync_group VSGF {
group {
LVS_F
}
}
vrrp_sync_group VSGB {
group {
LVS_B
}
}
But I saw that in my logs :
( I have changed the advertissement interval to 5sec )
Apr 27 18:39:50 bar Keepalived_vrrp: VRRP_Instance(LVS_F) Transition to
MASTER STATE
Apr 27 18:39:50 bar Keepalived_vrrp: VRRP_Group(VSGO) Syncing instances
to MASTER state
Apr 27 18:39:55 bar Keepalived_vrrp: VRRP_Instance(LVS_F) Entering
MASTER STATE
Apr 27 18:42:15 bar Keepalived_vrrp: VRRP_Instance(LVS_F) Received
higher prio advert
Apr 27 18:42:15 bar Keepalived_vrrp: VRRP_Instance(LVS_F) Entering
BACKUP STATE
Apr 27 18:42:15 bar Keepalived_vrrp: VRRP_Group(VSGO) Syncing instances
to BACKUP state
Apr 27 18:39:50 foo Keepalived_vrrp: VRRP_Instance(LVS_F) Received lower
prio advert, forcing new election
Apr 27 18:40:05 foo Keepalived_vrrp: VRRP_Instance(LVS_F) Received lower
prio advert, forcing new election
Apr 27 18:40:10 foo Keepalived_vrrp: VRRP_Instance(LVS_F) Received lower
prio advert, forcing new election
Apr 27 18:40:20 foo Keepalived_vrrp: VRRP_Instance(LVS_F) Received lower
prio advert, forcing new election
[...]
Apr 27 18:42:00 foo Keepalived_vrrp: VRRP_Instance(LVS_F) Received lower
prio advert, forcing new election
Apr 27 18:42:05 foo Keepalived_vrrp: VRRP_Instance(LVS_F) Received lower
prio advert, forcing new election
Apr 27 18:42:10 foo Keepalived_vrrp: VRRP_Instance(LVS_F) Received lower
prio advert, forcing new election
Between 2min30, the 2 directors havnt be able to comunicate each other (
no multicast packet send ) and "bar" become master on LVS_B and LVS_F ,
but "foo" was already master on LVS_F instance during this time.
Same problems of synchronisation , so ... any clue ?
Regards,
Nicolas
|