RE: speed up a logical replica setup

Hayato Kuroda (Fujitsu) <kuroda.hayato@fujitsu.com>

From: "Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com>

To: 'Euler Taveira' <euler@eulerto.com>, Alexander Lakhin <exclusion@gmail.com>

Cc: "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>, "Zhijie Hou (Fujitsu)" <houzj.fnst@fujitsu.com>, Amit Kapila <amit.kapila16@gmail.com>

Date: 2024-07-30T08:37:38Z

Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

pg_createsubscriber: Remove obsolete comment
- 71795d1cb41b 17.0 landed
- 1330843bb78e 18.0 landed
pg_createsubscriber: Fix an unpredictable recovery wait time.
- e5ba6a5ab62c 17.0 landed
- 03b08c8f5f3e 18.0 landed
Fix unstable test in 040_pg_createsubscriber.
- ae4e072bad5f 17.0 landed
- 9fd8b331dfe1 18.0 landed
Fix the testcase introduced in commit 81d20fbf7a.
- ae395f0f7edb 18.0 landed
- 14387ab06503 17.0 landed
Further weaken new pg_createsubscriber test on Windows.
- 55c309fc5b08 17.0 landed
- a1333ec048fb 18.0 landed
Temporarily(?) weaken new pg_createsubscriber test on Windows.
- 54508209178b 17.0 landed
Make pg_createsubscriber warn if publisher has two-phase commit enabled.
- 917754557cc0 17.0 landed
Make pg_createsubscriber more wary about quoting connection parameters.
- b3f5ccebd79d 17.0 landed
pg_createsubscriber: Remove failover replication slots on subscriber
- 81d20fbf7a03 17.0 landed
pg_createsubscriber: Remove replication slot check on primary
- b96391382626 17.0 landed
pg_createsubscriber: Only --recovery-timeout controls the end of recovery process
- 04c8634c0c4d 17.0 landed
pg_createsubscriber: creates a new logical replica from a standby server
- d44032d01463 17.0 landed
Add some const decorations
- 48018f1d8c12 17.0 landed
Add option force_initdb to PostgreSQL::Test::Cluster:init()
- ff9e1e764fcc 17.0 cited
Remove MSVC scripts
- 1301c80b2167 17.0 cited

Dear Euler,

I applied your patch and confirmed it could fix the issue [1].

METHOD
=======
1. shortened the snapshot interval to 3ms, and
```
-#define LOG_SNAPSHOT_INTERVAL_MS 15000
+#define LOG_SNAPSHOT_INTERVAL_MS 3
```
2. ran test 040_pg_createsubscriber.pl many times.
   Before patching, I could reproduce the failure when I ran less than 1hr.
   After patching, I ran more than 1hr but I could not reproduce.

Since this is a timing issue which I could not surely reproduce, I wanted others
to test it as well.

[1]: https://www.postgresql.org/message-id/68de6498-0449-a113-dd03-e198dded0bac%40gmail.com

Best regards,
Hayato Kuroda
FUJITSU LIMITED