pageserver: fix unlogged relations with sharding (#7454)

## Problem

- #7451 

INIT_FORKNUM blocks must be stored on shard 0 to enable including them
in basebackup.

This issue can be missed in simple tests because creating an unlogged
table isn't sufficient -- to repro I had to create an _index_ on an
unlogged table (then restart the endpoint).

Closes: #7451 

## Summary of changes

- Add a reproducer for the issue.
- Tweak the condition for `key_is_shard0` to include anything that isn't
a normal relation block _and_ any normal relation block whose forknum is
INIT_FORKNUM.
- To enable existing databases to recover from the issue, add a special
case that omits relations if they were stored on the wrong INITFORK.
This enables postgres to start and the user to drop the table and
recreate it.
This commit is contained in:
John Spray
2024-04-22 12:47:24 +01:00
committed by Joonas Koivunen
parent 35e9fb360b
commit 459bc479dc
3 changed files with 83 additions and 3 deletions

View File

@@ -13,7 +13,7 @@
use anyhow::{anyhow, bail, ensure, Context};
use bytes::{BufMut, Bytes, BytesMut};
use fail::fail_point;
use pageserver_api::key::{key_to_slru_block, Key};
use pageserver_api::key::{key_to_slru_block, rel_block_to_key, Key};
use postgres_ffi::pg_constants;
use std::fmt::Write as FmtWrite;
use std::time::SystemTime;
@@ -297,7 +297,20 @@ where
if rel.forknum == INIT_FORKNUM {
// I doubt we need _init fork itself, but having it at least
// serves as a marker relation is unlogged.
self.add_rel(rel, rel).await?;
if let Err(_e) = self.add_rel(rel, rel).await {
if self
.timeline
.get_shard_identity()
.is_key_buggy_forknum(&rel_block_to_key(rel, 0x0))
{
// Workaround https://github.com/neondatabase/neon/issues/7451 -- if we have an unlogged relation
// whose INIT_FORKNUM is not correctly on shard zero, then omit it in the basebackup. This allows
// postgres to start up. The relation won't work, but it will be possible to DROP TABLE on it and
// recreate.
tracing::warn!("Omitting relation {rel} for issue #7451: drop and recreate this unlogged relation");
continue;
}
};
self.add_rel(rel, rel.with_forknum(MAIN_FORKNUM)).await?;
continue;
}