|
Hi All,
Sorry for the off-list question, but perhaps someone has already met the
problem, and may know the instant solution...
I have configured keepalived for failover on LVS. Everything works well,
the master sends keepalived messages to the backup, the backup takes the
role of the master whenever it goes down, releases all IPs when it comes
back again, etc. But after 3 or 4 hours of operation the backup turns to
be master, while the master is still running, without any explicit log
message, or event. Both the original master, and the new master act as
masters, which is strange!
1. I use ipvs-1.0.6 with RedHat linux kernel 2.4.18.
2. I checked iptables rules, and multicast messages are allowed to arrive
and send. I checked the messages with tcpdump, and the original backup
sends the keepalived messages, when the problem occurs.
3. I checked the keepalived configs, and priorities are set to different
values (backup priorities are lower), the states are set to MASTER and
BACKUP respectively, and the VID names are different for the master and
for the backup.
4. 10 minutes before the bogus behavior some messages appeared in the
master's log file:
Oct 10 14:55:10 ldir2 Keepalived: VRRP_Instance(VI_13) Received lower prio
advert, forcing new election
Oct 10 14:55:10 ldir2 Keepalived: VRRP_Instance(VI_13) Sending gratuitous
ARP on eth0
Any help would be highly appreciated!
Regards,
Peter Stefan
|
