This is the scenario that causes the NAT Table to be over-flown with entries:
A service was configured not to be synchronized between cluster members.
In addition, the connections on this service undergo NAT.
Meaning, for each connection on this service through Active member of ClusterXL in HA mode:
- an entry will be added to Connections Table, but these entries will not be synchronized between cluster members (since the service was configured to be synchronized)
- an entry will be added to NAT Table, and these entries will be synchronized between cluster members (by design, NAT Table is always synchronized)
Then, either the Active cluster member is stopped (either powered off , or 'cpstop'), or fail-over occurs to other members.
The entries of not synchronized connections will remain in NAT Table on all other cluster members, and will not be deleted, because they do not appear in the Connection table on other cluster members.
Reason: by design, an entry about a connection in NAT Table is deleted only if the corresponding entry is deleted from the Connections Table.