Mailing List Archive: [lvs-users] Backup instance brings up VIPs and keeps them up

[lvs-users] Backup instance brings up VIPs and keeps them up

Oct 22, 2013, 11:08 AM

Post #1 of 3 (1634 views)

Hello,

I have the following problem. Occasionally, the backup keepalived instance
becomes master, brings up the virtual IPs, but then it fails bringing them
down after it realizes that another instance with higher prio number is
online.

Here are excerpts from the log files. everything happens within a second so
timestamps are omitted.

On the backup server:

Keepalived_vrrp: VRRP_Instance(INTIF) Transition to MASTER STATE
Keepalived_vrrp: VRRP_Group(VG1) Syncing instances to MASTER state
Keepalived_vrrp: VRRP_Instance(EXTIF) Transition to MASTER STATE
Keepalived_vrrp: VRRP_Instance(EXTIF) Entering MASTER STATE
Keepalived_vrrp: VRRP_Instance(INTIF) Received higher prio advert
Keepalived_vrrp: VRRP_Instance(INTIF) IPSEC-AH : Syncing seq_num -
Decrement seq
Keepalived_vrrp: VRRP_Instance(INTIF) Entering BACKUP STATE
Keepalived_vrrp: VRRP_Group(VG1) Syncing instances to BACKUP state
kernel: [31833.448553] IPVS: stopping backup sync thread 2564 ...
kernel: [31833.451613] IPVS: stopping master sync thread 2587 ...

On the master:

Keepalived_vrrp: VRRP_Instance(EXTIF) Received lower prio advert, forcing
new election
Keepalived_vrrp: VRRP_Instance(EXTIF) IPSEC-AH : Syncing seq_num -
Increment seq
Keepalived_vrrp: VRRP_Instance(INTIF) Received lower prio advert, forcing
new election
Keepalived_vrrp: VRRP_Instance(INTIF) IPSEC-AH : Syncing seq_num -
Increment seq
Keepalived_vrrp: VRRP_Instance(EXTIF) Received lower prio advert, forcing
new election
Keepalived_vrrp: VRRP_Instance(EXTIF) IPSEC-AH : Syncing seq_num -
Increment seq

INTIF and EXTIF are two vrrp instances associated with different interfaces
of the load balancer
VG1 is a vrrp_sync_group containing both INTIF and EXTIF vrrp instances.

auth_type is AH
advert_int is 1

In the case above, the VIPs associated with the EXTIF instance remained up
regardless that it was logged that the VG1 instances are syncing to backup
state. In other cases, the VIPs on both EXTIF and INTIF will remain up.

The environment is:
OS: Ubuntu 12.04.3 LTS
kernel: 3.2.0-54-generic
keepalived: 1.2.2-3ubuntu1.1
ipvsadm: 1.25.clean-1ubuntu5

Any idea what may be causing this?

Thanks!
_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/

LinuxVirtualServer.org mailing list - lvs-users@LinuxVirtualServer.org
Send requests to lvs-users-request@LinuxVirtualServer.org
or go to http://lists.graemef.net/mailman/listinfo/lvs-users

Re: [lvs-users] Backup instance brings up VIPs and keeps them up [ In reply to ]

rohara at redhat

Oct 22, 2013, 4:42 PM

Post #2 of 3 (1610 views)

Permalink

On Tue, Oct 22, 2013 at 11:08:58AM -0700, Stoyan Stoyanov wrote:
> Hello,
>
> I have the following problem. Occasionally, the backup keepalived instance
> becomes master, brings up the virtual IPs, but then it fails bringing them
> down after it realizes that another instance with higher prio number is
> online.
>
> Here are excerpts from the log files. everything happens within a second so
> timestamps are omitted.
>
> On the backup server:
>
> Keepalived_vrrp: VRRP_Instance(INTIF) Transition to MASTER STATE
> Keepalived_vrrp: VRRP_Group(VG1) Syncing instances to MASTER state
> Keepalived_vrrp: VRRP_Instance(EXTIF) Transition to MASTER STATE
> Keepalived_vrrp: VRRP_Instance(EXTIF) Entering MASTER STATE
> Keepalived_vrrp: VRRP_Instance(INTIF) Received higher prio advert
> Keepalived_vrrp: VRRP_Instance(INTIF) IPSEC-AH : Syncing seq_num -
> Decrement seq
> Keepalived_vrrp: VRRP_Instance(INTIF) Entering BACKUP STATE
> Keepalived_vrrp: VRRP_Group(VG1) Syncing instances to BACKUP state
> kernel: [31833.448553] IPVS: stopping backup sync thread 2564 ...
> kernel: [31833.451613] IPVS: stopping master sync thread 2587 ...
>
> On the master:
>
> Keepalived_vrrp: VRRP_Instance(EXTIF) Received lower prio advert, forcing
> new election
> Keepalived_vrrp: VRRP_Instance(EXTIF) IPSEC-AH : Syncing seq_num -
> Increment seq
> Keepalived_vrrp: VRRP_Instance(INTIF) Received lower prio advert, forcing
> new election
> Keepalived_vrrp: VRRP_Instance(INTIF) IPSEC-AH : Syncing seq_num -
> Increment seq
> Keepalived_vrrp: VRRP_Instance(EXTIF) Received lower prio advert, forcing
> new election
> Keepalived_vrrp: VRRP_Instance(EXTIF) IPSEC-AH : Syncing seq_num -
> Increment seq
>
> INTIF and EXTIF are two vrrp instances associated with different interfaces
> of the load balancer
> VG1 is a vrrp_sync_group containing both INTIF and EXTIF vrrp instances.
>
> auth_type is AH
> advert_int is 1
>
> In the case above, the VIPs associated with the EXTIF instance remained up
> regardless that it was logged that the VG1 instances are syncing to backup
> state. In other cases, the VIPs on both EXTIF and INTIF will remain up.
>
> The environment is:
> OS: Ubuntu 12.04.3 LTS
> kernel: 3.2.0-54-generic
> keepalived: 1.2.2-3ubuntu1.1
> ipvsadm: 1.25.clean-1ubuntu5
>
> Any idea what may be causing this?

Can you post your config file? Also, you might have better luck asking
on the keepalived-devel mailing list.

Ryan

_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/

LinuxVirtualServer.org mailing list - lvs-users@LinuxVirtualServer.org
Send requests to lvs-users-request@LinuxVirtualServer.org
or go to http://lists.graemef.net/mailman/listinfo/lvs-users

Re: [lvs-users] Backup instance brings up VIPs and keeps them up [ In reply to ]

stoyansbg at gmail

Oct 23, 2013, 9:08 AM

Post #3 of 3 (1616 views)

Permalink

The VRRP part of the keepalived configuration is:

vrrp_sync_group VG1 {
group {
INTIF
EXTIF
}
}

vrrp_instance INTIF {
state MASTER
interface eth1
virtual_router_id 51
priority 150
advert_int 1
lvs_sync_daemon_interface eth1
authentication {
auth_type AH
auth_pass somepass1
}

virtual_ipaddress {
192.168.50.120
}
}

vrrp_instance EXTIF {
state MASTER
interface eth0
virtual_router_id 52
priority 150
advert_int 1
lvs_sync_daemon_interface eth1
authentication {
auth_type AH
auth_pass somepass2
}

virtual_ipaddress {
192.168.100.161
192.168.100.162
192.168.100.163
}
}

The backup server config differs only by the state option being set to
BACKUP and priority set to 100.

On Tue, Oct 22, 2013 at 4:42 PM, Ryan O'Hara <rohara@redhat.com> wrote:

> On Tue, Oct 22, 2013 at 11:08:58AM -0700, Stoyan Stoyanov wrote:
> > Hello,
> >
> > I have the following problem. Occasionally, the backup keepalived
> instance
> > becomes master, brings up the virtual IPs, but then it fails bringing
> them
> > down after it realizes that another instance with higher prio number is
> > online.
> >
> > Here are excerpts from the log files. everything happens within a second
> so
> > timestamps are omitted.
> >
> > On the backup server:
> >
> > Keepalived_vrrp: VRRP_Instance(INTIF) Transition to MASTER STATE
> > Keepalived_vrrp: VRRP_Group(VG1) Syncing instances to MASTER state
> > Keepalived_vrrp: VRRP_Instance(EXTIF) Transition to MASTER STATE
> > Keepalived_vrrp: VRRP_Instance(EXTIF) Entering MASTER STATE
> > Keepalived_vrrp: VRRP_Instance(INTIF) Received higher prio advert
> > Keepalived_vrrp: VRRP_Instance(INTIF) IPSEC-AH : Syncing seq_num -
> > Decrement seq
> > Keepalived_vrrp: VRRP_Instance(INTIF) Entering BACKUP STATE
> > Keepalived_vrrp: VRRP_Group(VG1) Syncing instances to BACKUP state
> > kernel: [31833.448553] IPVS: stopping backup sync thread 2564 ...
> > kernel: [31833.451613] IPVS: stopping master sync thread 2587 ...
> >
> > On the master:
> >
> > Keepalived_vrrp: VRRP_Instance(EXTIF) Received lower prio advert, forcing
> > new election
> > Keepalived_vrrp: VRRP_Instance(EXTIF) IPSEC-AH : Syncing seq_num -
> > Increment seq
> > Keepalived_vrrp: VRRP_Instance(INTIF) Received lower prio advert, forcing
> > new election
> > Keepalived_vrrp: VRRP_Instance(INTIF) IPSEC-AH : Syncing seq_num -
> > Increment seq
> > Keepalived_vrrp: VRRP_Instance(EXTIF) Received lower prio advert, forcing
> > new election
> > Keepalived_vrrp: VRRP_Instance(EXTIF) IPSEC-AH : Syncing seq_num -
> > Increment seq
> >
> > INTIF and EXTIF are two vrrp instances associated with different
> interfaces
> > of the load balancer
> > VG1 is a vrrp_sync_group containing both INTIF and EXTIF vrrp instances.
> >
> > auth_type is AH
> > advert_int is 1
> >
> > In the case above, the VIPs associated with the EXTIF instance remained
> up
> > regardless that it was logged that the VG1 instances are syncing to
> backup
> > state. In other cases, the VIPs on both EXTIF and INTIF will remain up.
> >
> > The environment is:
> > OS: Ubuntu 12.04.3 LTS
> > kernel: 3.2.0-54-generic
> > keepalived: 1.2.2-3ubuntu1.1
> > ipvsadm: 1.25.clean-1ubuntu5
> >
> > Any idea what may be causing this?
>
> Can you post your config file? Also, you might have better luck asking
> on the keepalived-devel mailing list.
>
> Ryan
>
>
> _______________________________________________
> Please read the documentation before posting - it's available at:
> http://www.linuxvirtualserver.org/
>
> LinuxVirtualServer.org mailing list - lvs-users@LinuxVirtualServer.org
> Send requests to lvs-users-request@LinuxVirtualServer.org
> or go to http://lists.graemef.net/mailman/listinfo/lvs-users
>
_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/

LinuxVirtualServer.org mailing list - lvs-users@LinuxVirtualServer.org
Send requests to lvs-users-request@LinuxVirtualServer.org
or go to http://lists.graemef.net/mailman/listinfo/lvs-users