Dear Fujii-san,

Sorry for the late reply, I missed your post.

> +$node->wait_for_log(
> + qr/START_REPLICATION SLOT \"reconnect_test\" LOGICAL/,
> + $log_offset);
> 
> Could this pattern also match the STATEMENT log entry emitted when
> the old walsender is terminated by pg_terminate_backend()?

Good catch, it can be.

> If so,
> is it be better to wait for the new slot acquisition instead? For example:
> 
>     qr/acquired logical replication slot \"reconnect_test\"/,

Another way is to check the start position specified by the START_REPLICATION is
advanced, but your approach seems better. PSA new version.

Best regards,
Hayato Kuroda
FUJITSU LIMITED

Attachment: v2-0001-Make-030_pg_recvlogical-robust.patch
Description: v2-0001-Make-030_pg_recvlogical-robust.patch

Reply via email to