opensourcemirror
/
postgresql
mirror of git://git.postgresql.org/git/postgresql.git
1
0
Fork 0
Code Issues Releases Wiki Activity

Fix the review comments and a bug in the slot sync code. 

Ensure that when updating the catalog_xmin of the synced slots, it is
first written to disk before changing the in-memory value
(effective_catalog_xmin). This is to prevent a scenario where the
in-memory value change triggers a vacuum to remove catalog tuples before
the catalog_xmin is written to disk. In the event of a crash before the
catalog_xmin is persisted, we would not know that some required catalog
tuples have been removed and the synced slot would be invalidated.

Change the sanity check to ensure that remote_slot's confirmed_flush LSN
can't precede the local/synced slot during slot sync. Note that the
restart_lsn of the synced/local slot can be ahead of remote_slot. This can
happen when slot advancing machinery finds a running xacts record after
reaching the consistent state at a later point than the primary where it
serializes the snapshot and updates the restart_lsn.

Make the check to sync slots robust by allowing to sync only when the
confirmed_lsn, restart_lsn, or catalog_xmin of the remote slot is ahead of
the synced/local slot.

Reported-by: Amit Kapila and Shveta Malik
Author: Hou Zhijie, Shveta Malik
Reviewed-by: Amit Kapila, Bertrand Drouvot
Discussion: https://postgr.es/m/OS0PR01MB57162B67D3CB01B2756FBA6D94062@OS0PR01MB5716.jpnprd01.prod.outlook.com
Discussion: https://postgr.es/m/CAJpy0uCSS5zmdyUXhvw41HSdTbRqX1hbYqkOfHNj7qQ+2zn0AQ@mail.gmail.com
This commit is contained in:
Amit Kapila 2024-04-12 15:03:28 +05:30
parent 3af7040985
commit 3741f2a09d
1 changed files with 118 additions and 43 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

161
src/backend/replication/logical/slotsync.c
View File

@ -162,22 +162,78 @@ static void update_synced_slots_inactive_since(void);
* *found_consistent_snapshot will be true iff the remote slot's LSN or xmin is
* modified, and decoding from the corresponding LSN's can reach a
* consistent snapshot.
*
* *remote_slot_precedes will be true if the remote slot's LSN or xmin
* precedes locally reserved position.
*/
static bool
update_local_synced_slot(RemoteSlot *remote_slot, Oid remote_dbid,
bool *found_consistent_snapshot)
bool *found_consistent_snapshot,
bool *remote_slot_precedes)
{
ReplicationSlot *slot = MyReplicationSlot;
bool slot_updated = false;
bool updated_xmin_or_lsn = false;
bool updated_config = false;
Assert(slot->data.invalidated == RS_INVAL_NONE);
if (found_consistent_snapshot)
*found_consistent_snapshot = false;
if (remote_slot->confirmed_lsn != slot->data.confirmed_flush ||
remote_slot->restart_lsn != slot->data.restart_lsn ||
remote_slot->catalog_xmin != slot->data.catalog_xmin)
if (remote_slot_precedes)
*remote_slot_precedes = false;
/*
* Don't overwrite if we already have a newer catalog_xmin and
* restart_lsn.
*/
if (remote_slot->restart_lsn < slot->data.restart_lsn ||
TransactionIdPrecedes(remote_slot->catalog_xmin,
slot->data.catalog_xmin))
{
/*
* This can happen in following situations:
*
* If the slot is temporary, it means either the initial WAL location
* reserved for the local slot is ahead of the remote slot's
* restart_lsn or the initial xmin_horizon computed for the local slot
* is ahead of the remote slot.
*
* If the slot is persistent, restart_lsn of the synced slot could
* still be ahead of the remote slot. Since we use slot advance
* functionality to keep snapbuild/slot updated, it is possible that
* the restart_lsn is advanced to a later position than it has on the
* primary. This can happen when slot advancing machinery finds
* running xacts record after reaching the consistent state at a later
* point than the primary where it serializes the snapshot and updates
* the restart_lsn.
*
* We LOG the message if the slot is temporary as it can help the user
* to understand why the slot is not sync-ready. In the case of a
* persistent slot, it would be a more common case and won't directly
* impact the users, so we used DEBUG1 level to log the message.
*/
ereport(slot->data.persistency == RS_TEMPORARY ? LOG : DEBUG1,
errmsg("could not sync slot \"%s\" as remote slot precedes local slot",
remote_slot->name),
errdetail("Remote slot has LSN %X/%X and catalog xmin %u, but local slot has LSN %X/%X and catalog xmin %u.",
LSN_FORMAT_ARGS(remote_slot->restart_lsn),
remote_slot->catalog_xmin,
LSN_FORMAT_ARGS(slot->data.restart_lsn),
slot->data.catalog_xmin));
if (remote_slot_precedes)
*remote_slot_precedes = true;
}
/*
* Attempt to sync LSNs and xmins only if remote slot is ahead of local
* slot.
*/
else if (remote_slot->confirmed_lsn > slot->data.confirmed_flush ||
remote_slot->restart_lsn > slot->data.restart_lsn ||
TransactionIdFollows(remote_slot->catalog_xmin,
slot->data.catalog_xmin))
{
/*
* We can't directly copy the remote slot's LSN or xmin unless there
@ -198,7 +254,6 @@ update_local_synced_slot(RemoteSlot *remote_slot, Oid remote_dbid,
slot->data.restart_lsn = remote_slot->restart_lsn;
slot->data.confirmed_flush = remote_slot->confirmed_lsn;
slot->data.catalog_xmin = remote_slot->catalog_xmin;
slot->effective_catalog_xmin = remote_slot->catalog_xmin;
SpinLockRelease(&slot->mutex);
if (found_consistent_snapshot)
@ -208,12 +263,18 @@ update_local_synced_slot(RemoteSlot *remote_slot, Oid remote_dbid,
{
LogicalSlotAdvanceAndCheckSnapState(remote_slot->confirmed_lsn,
found_consistent_snapshot);
/* Sanity check */
if (slot->data.confirmed_flush != remote_slot->confirmed_lsn)
ereport(ERROR,
errmsg_internal("synchronized confirmed_flush for slot \"%s\" differs from remote slot",
remote_slot->name),
errdetail_internal("Remote slot has LSN %X/%X but local slot has LSN %X/%X.",
LSN_FORMAT_ARGS(remote_slot->confirmed_lsn),
LSN_FORMAT_ARGS(slot->data.confirmed_flush)));
}
ReplicationSlotsComputeRequiredXmin(false);
ReplicationSlotsComputeRequiredLSN();
slot_updated = true;
updated_xmin_or_lsn = true;
}
if (remote_dbid != slot->data.database ||
@ -233,10 +294,37 @@ update_local_synced_slot(RemoteSlot *remote_slot, Oid remote_dbid,
slot->data.failover = remote_slot->failover;
SpinLockRelease(&slot->mutex);
slot_updated = true;
updated_config = true;
}
return slot_updated;
/*
* We have to write the changed xmin to disk *before* we change the
* in-memory value, otherwise after a crash we wouldn't know that some
* catalog tuples might have been removed already.
*/
if (updated_config || updated_xmin_or_lsn)
{
ReplicationSlotMarkDirty();
ReplicationSlotSave();
}
/*
* Now the new xmin is safely on disk, we can let the global value
* advance. We do not take ProcArrayLock or similar since we only advance
* xmin here and there's not much harm done by a concurrent computation
* missing that.
*/
if (updated_xmin_or_lsn)
{
SpinLockAcquire(&slot->mutex);
slot->effective_catalog_xmin = remote_slot->catalog_xmin;
SpinLockRelease(&slot->mutex);
ReplicationSlotsComputeRequiredXmin(false);
ReplicationSlotsComputeRequiredLSN();
}
return updated_config || updated_xmin_or_lsn;
}
/*
@ -460,14 +548,17 @@ update_and_persist_local_synced_slot(RemoteSlot *remote_slot, Oid remote_dbid)
{
ReplicationSlot *slot = MyReplicationSlot;
bool found_consistent_snapshot = false;
bool remote_slot_precedes = false;
(void) update_local_synced_slot(remote_slot, remote_dbid,
&found_consistent_snapshot,
&remote_slot_precedes);
/*
* Check if the primary server has caught up. Refer to the comment atop
* the file for details on this check.
*/
if (remote_slot->restart_lsn < slot->data.restart_lsn ||
TransactionIdPrecedes(remote_slot->catalog_xmin,
slot->data.catalog_xmin))
if (remote_slot_precedes)
{
/*
* The remote slot didn't catch up to locally reserved position.
@ -476,23 +567,10 @@ update_and_persist_local_synced_slot(RemoteSlot *remote_slot, Oid remote_dbid)
* current location when recreating the slot in the next cycle. It may
* take more time to create such a slot. Therefore, we keep this slot
* and attempt the synchronization in the next cycle.
*
* XXX should this be changed to elog(DEBUG1) perhaps?
*/
ereport(LOG,
errmsg("could not sync slot \"%s\" as remote slot precedes local slot",
remote_slot->name),
errdetail("Remote slot has LSN %X/%X and catalog xmin %u, but local slot has LSN %X/%X and catalog xmin %u.",
LSN_FORMAT_ARGS(remote_slot->restart_lsn),
remote_slot->catalog_xmin,
LSN_FORMAT_ARGS(slot->data.restart_lsn),
slot->data.catalog_xmin));
return false;
}
(void) update_local_synced_slot(remote_slot, remote_dbid,
&found_consistent_snapshot);
/*
* Don't persist the slot if it cannot reach the consistent point from the
* restart_lsn. See comments atop this file.
@ -633,23 +711,20 @@ synchronize_one_slot(RemoteSlot *remote_slot, Oid remote_dbid)
/*
* Sanity check: As long as the invalidations are handled
* appropriately as above, this should never happen.
*
* We don't need to check restart_lsn here. See the comments in
* update_local_synced_slot() for details.
*/
if (remote_slot->restart_lsn < slot->data.restart_lsn)
elog(ERROR,
"cannot synchronize local slot \"%s\" LSN(%X/%X)"
" to remote slot's LSN(%X/%X) as synchronization"
" would move it backwards", remote_slot->name,
LSN_FORMAT_ARGS(slot->data.restart_lsn),
LSN_FORMAT_ARGS(remote_slot->restart_lsn));
if (remote_slot->confirmed_lsn < slot->data.confirmed_flush)
ereport(ERROR,
errmsg_internal("cannot synchronize local slot \"%s\"",
remote_slot->name),
errdetail_internal("Local slot's start streaming location LSN(%X/%X) is ahead of remote slot's LSN(%X/%X).",
LSN_FORMAT_ARGS(slot->data.confirmed_flush),
LSN_FORMAT_ARGS(remote_slot->confirmed_lsn)));
/* Make sure the slot changes persist across server restart */
if (update_local_synced_slot(remote_slot, remote_dbid, NULL))
{
ReplicationSlotMarkDirty();
ReplicationSlotSave();
slot_updated = true;
}
slot_updated = update_local_synced_slot(remote_slot, remote_dbid,
NULL, NULL);
}
}
/* Otherwise create the slot first. */