The step to reproduce this issue.
1. Create a table
2. Insert data
insert into gist_point_tbl (id, p) select g, point(g*10, g*10) from generate_series(1, 1000000) g;
3. Delete data
4. Vacuum table
-- Send SIGINT to vacuum process after WAL-log of the truncation is flushed and the truncation is not finished
-- We will receive error message "ERROR: canceling statement due to user request"
5. Vacuum table again
-- The standby node crashed and the PANIC log is "PANIC: WAL contains references to invalid pages"
The standby node succeed to replay truncate log but master node doesn't truncate the file, it will be crashed if master node writes to these blocks which truncated in standby node.
I try to fix issue to prevent query cancel interrupts during truncating.
diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c
index 5df4382b7e..04b696ae01 100644
--- a/src/backend/catalog/storage.c
+++ b/src/backend/catalog/storage.c
@@ -26,6 +26,7 @@
#include "access/xlogutils.h"
#include "catalog/storage.h"
#include "catalog/storage_xlog.h"
+#include "miscadmin.h"
#include "storage/freespace.h"
#include "storage/smgr.h"
#include "utils/memutils.h"
@@ -248,6 +249,14 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
if (vm)
visibilitymap_truncate(rel, nblocks);
+ /*
+ * When master node flush WAL-log of the truncation and then receive SIGINT signal to cancel
+ * this transaction before the truncation, if standby receive this WAL-log and do the truncation,
+ * standby node will crash when master node writes to these blocks which are truncated in standby node.
+ * So we prevent query cancel interrupts.
+ */
+ HOLD_CANCEL_INTERRUPTS();
+
/*
* We WAL-log the truncation before actually truncating, which means
* trouble if the truncation fails. If we then crash, the WAL replay
@@ -288,6 +297,8 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
/* Do the real work */
smgrtruncate(rel->rd_smgr, MAIN_FORKNUM, nblocks);
+
+ RESUME_CANCEL_INTERRUPTS();
}
