pageserver: revisit pageserver_wal_ingest_clear_vm_bits_unknown

[erikgrinaker]

We don't seem to be hitting this in production after #9911. Revisit in a few weeks time to tighten up the logic to error instead.

neon/pageserver/src/walingest.rs

Lines 330 to 365 in cc37fa0

	// Sometimes, Postgres seems to create heap WAL records with the
	// ALL_VISIBLE_CLEARED flag set, even though the bit in the VM page is
	// not set. In fact, it's possible that the VM page does not exist at all.
	// In that case, we don't want to store a record to clear the VM bit;
	// replaying it would fail to find the previous image of the page, because
	// it doesn't exist. So check if the VM page(s) exist, and skip the WAL
	// record if it doesn't.
	//
	// TODO: analyze the metrics and tighten this up accordingly. This logic
	// implicitly assumes that VM pages see explicit WAL writes before
	// implicit ClearVmBits, and will otherwise silently drop updates.
	let Some(vm_size) = get_relsize(modification, vm_rel, ctx).await? else {
	WAL_INGEST
	.clear_vm_bits_unknown
	.with_label_values(&["relation"])
	.inc();
	return Ok(());
	};
	if let Some(blknum) = new_vm_blk {
	if blknum >= vm_size {
	WAL_INGEST
	.clear_vm_bits_unknown
	.with_label_values(&["new_page"])
	.inc();
	new_vm_blk = None;
	}
	}
	if let Some(blknum) = old_vm_blk {
	if blknum >= vm_size {
	WAL_INGEST
	.clear_vm_bits_unknown
	.with_label_values(&["old_page"])
	.inc();
	old_vm_blk = None;
	}
	}