But PostgreSQL's logical decoding requires a contiguous WAL stream to maintain a valid catalog_xmin and restart_lsn, so it'll still fail to advance. So your replication from that node is broken, and you have to part the node then rejoin. You'll need to manually recover any diverged data.
Yup, I willing to reconcile data manualy via dblink or else,
is it still possible to move wal segment in 0.9.3?
Well, you can skip changes on the logical decoding stream manually in 0.9.3 using the underlying postgres functions. But it won't help you because what you broke when you deleted the WAL segments wasn't BDR, it was postgres logical decoding.
It *requires* a contiguous stream of WAL. It cannot recover if you do not have that. And you cannot really reset it.
You could drop and re-create the replication slot, I guess. But your nodes would be hopelessly out of sync and need manual resync (with data replication disabled) of one node vs another.
