Use transaction start/cancel for persistent writes to avoid leaving the database in an inconsistent state if we crash during the operation

Signed-off-by: Ronnie Sahlberg <ronniesahlberg@gmail.com>
(This used to be commit 09329f1f9114af44fc4e5e4f29a7315912313125)

3 files changed, 48 insertions, 6 deletions

diff --git a/source3/include/ctdbd_conn.h b/source3/include/ctdbd_conn.h
index 6e1b2f737a..3ea895d133 100644
--- a/source3/include/ctdbd_conn.h
+++ b/source3/include/ctdbd_conn.h
@@ -66,5 +66,7 @@ NTSTATUS ctdbd_register_ips(struct ctdbd_connection *conn,
 NTSTATUS ctdbd_register_reconfigure(struct ctdbd_connection *conn);
 
 NTSTATUS ctdbd_persistent_store(struct ctdbd_connection *conn, uint32_t db_id, TDB_DATA key, TDB_DATA data);
+NTSTATUS ctdbd_start_persistent_update(struct ctdbd_connection *conn, uint32_t db_id, TDB_DATA key, TDB_DATA data);
+NTSTATUS ctdbd_cancel_persistent_update(struct ctdbd_connection *conn, uint32_t db_id, TDB_DATA key, TDB_DATA data);
 
 #endif /* _CTDBD_CONN_H */
diff --git a/source3/lib/ctdbd_conn.c b/source3/lib/ctdbd_conn.c
index e8d3fd1159..51fefa93d4 100644
--- a/source3/lib/ctdbd_conn.c
+++ b/source3/lib/ctdbd_conn.c
@@ -1207,9 +1207,9 @@ NTSTATUS ctdbd_register_reconfigure(struct ctdbd_connection *conn)
 }
 
 /*
-  persstent store. Used when we update a record in a persistent database
+  persistent call. Used to start, store or cancel persistent updates.
  */
-NTSTATUS ctdbd_persistent_store(struct ctdbd_connection *conn, uint32_t db_id, TDB_DATA key, TDB_DATA data)
+static NTSTATUS ctdbd_persistent_call(struct ctdbd_connection *conn, uint32_t opcode, uint32_t db_id, TDB_DATA key, TDB_DATA data)
 {
        int cstatus=0;
        struct ctdb_rec_data *rec;
@@ -1232,8 +1232,7 @@ NTSTATUS ctdbd_persistent_store(struct ctdbd_connection *conn, uint32_t db_id, T
        recdata.dptr  = (uint8_t *)rec;
        recdata.dsize = length;
 
-       status = ctdbd_control(conn, CTDB_CURRENT_NODE, 
-                              CTDB_CONTROL_PERSISTENT_STORE, 
+       status = ctdbd_control(conn, CTDB_CURRENT_NODE, opcode,
                               0, recdata, NULL, NULL, &cstatus);
        if (cstatus != 0) {
                return NT_STATUS_INTERNAL_DB_CORRUPTION;
@@ -1241,6 +1240,40 @@ NTSTATUS ctdbd_persistent_store(struct ctdbd_connection *conn, uint32_t db_id, T
        return status;
 }
 
+/*
+  persistent store. Used when we update a record in a persistent database
+ */
+NTSTATUS ctdbd_persistent_store(struct ctdbd_connection *conn, uint32_t db_id, TDB_DATA key, TDB_DATA data)
+{
+	return ctdbd_persistent_call(conn,
+				CTDB_CONTROL_PERSISTENT_STORE, 
+				db_id, key, data);
+}
+
+/*
+  tell the ctdb daemon that we are starting a persistent update operation.
+  If we terminate/disconnect from the daemon without first performing
+  either a persistent_store or a cancel ctdbd will perform recovery.
+ */
+NTSTATUS ctdbd_start_persistent_update(struct ctdbd_connection *conn, uint32_t db_id, TDB_DATA key, TDB_DATA data)
+{
+	return ctdbd_persistent_call(conn,
+				CTDB_CONTROL_START_PERSISTENT_UPDATE, 
+				db_id, key, data);
+                              
+}
+
+/*
+  Cancel a persistent update operation. This is used if we have started a
+  persistent update  but we want to abort it before we have made changes to
+  the tdb database.
+ */
+NTSTATUS ctdbd_cancel_persistent_update(struct ctdbd_connection *conn, uint32_t db_id, TDB_DATA key, TDB_DATA data)
+{
+	return ctdbd_persistent_call(conn,
+				CTDB_CONTROL_CANCEL_PERSISTENT_UPDATE, 
+				db_id, key, data);
+}
 
 #else
 
diff --git a/source3/lib/dbwrap_ctdb.c b/source3/lib/dbwrap_ctdb.c
index cb4c573eb0..aa2f9d371c 100644
--- a/source3/lib/dbwrap_ctdb.c
+++ b/source3/lib/dbwrap_ctdb.c
@@ -79,12 +79,19 @@ static NTSTATUS db_ctdb_store_persistent(struct db_record *rec, TDB_DATA data, i
 	memcpy(cdata.dptr, &crec->header, sizeof(crec->header));
 	memcpy(cdata.dptr + sizeof(crec->header), data.dptr, data.dsize);
 
-	ret = tdb_store(crec->ctdb_ctx->wtdb->tdb, rec->key, cdata, TDB_REPLACE);
-	status = (ret == 0) ? NT_STATUS_OK : NT_STATUS_INTERNAL_DB_CORRUPTION;
+	status = ctdbd_start_persistent_update(messaging_ctdbd_connection(), crec->ctdb_ctx->db_id, rec->key, cdata);
 	
 	/* now tell ctdbd to update this record on all other nodes */
 	if (NT_STATUS_IS_OK(status)) {
+		ret = tdb_store(crec->ctdb_ctx->wtdb->tdb, rec->key, cdata, TDB_REPLACE);
+		status = (ret == 0) ? NT_STATUS_OK : NT_STATUS_INTERNAL_DB_CORRUPTION;
+	}
+
+	/* now tell ctdbd to update this record on all other nodes */
+	if (NT_STATUS_IS_OK(status)) {
 		status = ctdbd_persistent_store(messaging_ctdbd_connection(), crec->ctdb_ctx->db_id, rec->key, cdata);
+	} else {
+		ctdbd_cancel_persistent_update(messaging_ctdbd_connection(), crec->ctdb_ctx->db_id, rec->key, cdata);
 	}
 
 	SAFE_FREE(cdata.dptr);