Problem:
I am trying to test golden gate failover without application VIP.
1. Having two separate RACs.
2. EXT1 parameters on source RAC:
EXTRACT ext1
USERID ggcw, PASSWORD ggcw
RMTHOST gghost.example.com, MGRPORT 7809
RMTTRAIL ./dirdat/rt
TABLE marik.table1;
3. REP1 parameters on the target RAC:
REPLICAT rep1
ASSUMETARGETDEFS
USERID ggcw, PASSWORD ggcw
MAP marik.table1, TARGET marik.table1;
4. /etc/racdns entries on the source cluster:
10.100.0.7 gghost.example.com
10.100.0.9 gghost.example.com
Where 10.100.0.7 and 10.100.0.9 are nodes' IP addresses on the target cluster.
5. Target node listens incoming connection on 0 0.0.0.0:7809
tcp 0 0 0.0.0.0:7809 0.0.0.0:* LISTEN 43626/./mgr
6. EXT1 is able to connect to the target and sends replication data if it is able to resolve correct IP (where golden gate is started), but if I relocate golden gate service to another node, then EXT1 stops working and after each AUTORESTART connection is tried to the previous IP (where gg service is not running anymore ). So EXT1 fails with connection refused error:
2019-02-20T06:04:02.918+0000 ERROR OGG-01224 Oracle GoldenGate Capture for Oracle, ext1.prm: TCP/IP error 111 (Connection refused), endpoint: gghost.example.com:7809.
2019-02-20T06:04:02.918+0000 ERROR OGG-01668 Oracle GoldenGate Capture for Oracle, ext1.prm: PROCESS ABENDING.
In my opinion EXT1 (golden gate on source) should resolve gghost.example.com every time it tries to start and DNS should return this address in round robin fashion. By that way EXT1 should be able to connect to the right host maximum for the 2nd try.