Hi,
We're migrating from 9.6 to 14, using pglogical. We have several logical slots on the 9.6 instance implementing a change data capture pattern. For the migration, we plan on recreating the slots on the 14 instance, without taking a snapshot of the data. When the migration happens, we will simply start using the slots on the 14 instance (with the understanding that the LSNs won't match between the 9.6 and 14 instances).
In testing, we have this working, but there was a wrinkle and I'd like to know if my understanding is correct. On the 9.6 instance, when creating the replication slots, we would use the START_REPLICATION SLOT command, and then immediately take a snapshot, and it has worked great for years.
On a 14 instance, receiving changes from the 9.6 instance via pglogical, if we recreate the logical slot, but not take a snapshot, no changes are propagated, and there are no errors. However, if we first issue a 'CREATE PUBLICATION' command, then changes are propagated to the slot as expected.
We never issued a 'CREATE PUBLICATION' command on the 9.6 instance. My guess is that the act of taking a snapshot twittled whatever bits were necessary to propagate changes through the slot. By not taking a snapshot on the 14 instance, that doesn't happen, hence the need for the 'CREATE PUBLICATION' command. Alternatively, something was changed between 9.6 and 14 that now requires the new command (I looked through the docs and could find no mention, however).
Is my understanding correct? Am I missing something? I just want to make sure I'm not screwing something up.
Thanks,
Mark