POC: MZ reading from physical standby (DNM)#36923
Draft
martykulma wants to merge 1 commit into
Draft
Conversation
peterdukelarsen
added a commit
that referenced
this pull request
Jun 18, 2026
…#37020) ### Motivation We would like to be able to set up replication against a physical Postgres replica. This fails hard today because you can't read pg_current_wal_lsn from a standby. Relevant issue: https://linear.app/materializeinc/issue/SS-187/replication-from-postgres-replica This is adapted from #36923. ### Description - Record during purification whether the connection is to a physical replica - Use the appropriate method of loading the LSN depending on whether it's a physical replica or not ### Tradeoffs to call out - Replicating from a primary without `hot_standby_feedback=on` on the replica risks snapshots timing out (potentially quickly) and increases the risk of the replication slot being terminated. - In the test case where we stall, we'll hold a slot open on the replica. With `hot_standby_feedback=on` this risks bloating the disk of the primary. I think setting max_slot_wal_keep_size to a reasonably high level mitigates the scariest versions of this. This style of tradeoff seems somewhat similar to existing trade-offs around max_slot_wall_keep_size settings. - RTR is no longer accurate relative to the primary postgres replica. This should be ok and similar to cascading logical replication which we support today. - Materialize sources will require a fresh snapshot in the case of failover -- this is similar to logical replication today where a timeline id change produces an error. This protects us from corruption. ### Verification - Testdrive tests covering reading from a replica, basic RTR queries, and ensuring we stall if the same replica is recovered from a more recent backup
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
A proof of concept of MZ reading from a physical standby (PG 16 added logical decoding from physical standby). This doesn't account for what happens on a primary/secondary failover (my understanding is that the replication slot will be lost and the PG timeline will change, so the source would need to be recreated in MZ).