Symptoms
Customer performed NG load balancer restart during maintenance. After LB went online, all sites cannot load: despite 200 OK status is being returned, no content is displayed in browser.
In the load balancer's messages
log the following entries could be found:
pulse: SIOCGIFADDR failed: Cannot assign requested address
pulse: SIOCGIFADDR failed: Cannot assign requested address
pulse[6988]: We are backup node and backup is marked inactive -- exiting pulse
Pulse service is not started:
# service pulse status
pulse dead but subsys locked
Cause
Internal (backnet) address of Load Balancer was lost after reboot. Pulse finds discrepancy between actual network configuration and one defined in config file /etc/sysconfig/ha/lvs.cf
and interrupts startup.
Without pulse service, LB does not function at all as pulse is responsible for starting other LVS mechanisms: lvsd, nanny, ipvsadm
.
Resolution
Check the primary
address defined in /etc/sysconfig/ha/lvs.cf
:
primary = 10.10.20.30
Configure this address on one of interfaces of LB or pass this task to local network administrator. Ensure that Web nodes of cluster are reachable via backnet. Then you may start pulse service:
# service pulse start
And check that load balancer started to route requests:
# ipvsadm -ln
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
-> RemoteAddress:Port Forward Weight ActiveConn InActConn
FWM 100 wlc persistent 20 mask 255.255.255.0
-> 95.250.192.1:0 Route 320 110 48
-> 95.250.192.2:0 Route 160 61 15
-> 95.250.192.3:0 Route 320 108 52