Re: [PoC] pg_upgrade: allow to upgrade publisher node
Peter Smith <smithpb2250@gmail.com>
From: Peter Smith <smithpb2250@gmail.com>
To: "Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com>
Cc: Amit Kapila <amit.kapila16@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>,
Bruce Momjian <bruce@momjian.us>, Julien Rouhaud <rjuju123@gmail.com>, vignesh C <vignesh21@gmail.com>, Masahiko Sawada <sawada.mshk@gmail.com>,
"Zhijie Hou (Fujitsu)" <houzj.fnst@fujitsu.com>
Date: 2023-09-01T07:49:21Z
Lists: pgsql-hackers
Here are some review comments for v29-0002 ====== src/bin/pg_upgrade/check.c 1. check_old_cluster_for_valid_slots + /* Quick exit if the cluster does not have logical slots. */ + if (count_old_cluster_logical_slots() == 0) + return; /Quick exit/Quick return/ I know they are kind of the same, but the reason I previously suggested this change was to keep it consistent with the similar comment that is already in check_new_cluster_logical_replication_slots(). ~~~ 2. check_old_cluster_for_valid_slots + /* + * Do additional checks to ensure that confirmed_flush LSN of all the slots + * is the same as the latest checkpoint location. + * + * Note: This can be satisfied only when the old cluster has been shut + * down, so we skip this live checks. + */ + if (!live_check) missing word /skip this live checks./skip this for live checks./ ====== src/bin/pg_upgrade/controldata.c 3. + /* + * Read the latest checkpoint location if the cluster is PG17 + * or later. This is used for upgrading logical replication + * slots. Currently, we need it only for the old cluster but + * didn't add additional check for the similicity. + */ + if (GET_MAJOR_VERSION(cluster->major_version) >= 1700) /similicity/simplicity/ SUGGESTION Currently, we need it only for the old cluster but for simplicity chose not to have additional checks. ====== src/bin/pg_upgrade/info.c 4. get_old_cluster_logical_slot_infos_per_db + /* + * The temporary slots are expressly ignored while checking because such + * slots cannot exist after the upgrade. During the upgrade, clusters are + * started and stopped several times so that temporary slots will be + * removed. + */ + res = executeQueryOrDie(conn, "SELECT slot_name, plugin, two_phase " + "FROM pg_catalog.pg_replication_slots " + "WHERE slot_type = 'logical' AND " + "wal_status <> 'lost' AND " + "database = current_database() AND " + "temporary IS FALSE;"); IIUC, the removal of temp slots is just a side-effect of the start/stop; not the *reason* for the start/stop. So, the last sentence needs some modification BEFORE During the upgrade, clusters are started and stopped several times so that temporary slots will be removed. SUGGESTION During the upgrade, clusters are started and stopped several times causing any temporary slots to be removed. ------ Kind Regards, Peter Smith. Fujitsu Australia
Commits
-
Fix issues in binary_upgrade_logical_slot_has_caught_up().
- 0bf62460bb9e 17.0 landed
-
Fix a random failure in 003_logical_slots.pl.
- 63c5df126abb 17.0 landed
-
Fix a test in 003_logical_slots.
- 3e36e48d8efe 17.0 landed
-
Fix uninitialized slot array access during the upgrade.
- a7db71ed2787 17.0 landed
-
Fix the test 003_logical_slots.
- 8af917be6bad 17.0 landed
-
Commit b195e6d482 forgot to update meson.build.
- 8949b978ff49 17.0 landed
-
Use shorter file names in the upgrade logical slots test.
- b195e6d482b8 17.0 landed
-
Migrate logical slots to the new node during an upgrade.
- 29d0a77fa660 17.0 landed
-
Flush logical slots to disk during a shutdown checkpoint if required.
- e0b2eed047df 17.0 landed
-
Prevent possibility of panics during shutdown checkpoint.
- c6c333436491 10.0 cited