Fail fast on constraint violations in router executor

pull/552/head
Metin Doslu 2016-06-07 17:57:42 +03:00
parent 15eed396b3
commit 7d0c90b398
11 changed files with 113 additions and 69 deletions

View File

@ -827,7 +827,7 @@ MasterPartitionMethod(RangeVar *relation)
}
else
{
ReportRemoteError(masterConnection, queryResult);
ReportRemoteError(masterConnection, queryResult, false);
ereport(ERROR, (errmsg("could not get the partition method of the "
"distributed table")));
}
@ -924,7 +924,7 @@ OpenCopyTransactions(CopyStmt *copyStatement, ShardConnections *shardConnections
result = PQexec(connection, "BEGIN");
if (PQresultStatus(result) != PGRES_COMMAND_OK)
{
ReportRemoteError(connection, result);
ReportRemoteError(connection, result, false);
failedPlacementList = lappend(failedPlacementList, placement);
PQclear(result);
@ -937,7 +937,7 @@ OpenCopyTransactions(CopyStmt *copyStatement, ShardConnections *shardConnections
result = PQexec(connection, copyCommand->data);
if (PQresultStatus(result) != PGRES_COPY_IN)
{
ReportRemoteError(connection, result);
ReportRemoteError(connection, result, false);
failedPlacementList = lappend(failedPlacementList, placement);
PQclear(result);
@ -1504,7 +1504,7 @@ RemoteCreateEmptyShard(char *relationName)
}
else
{
ReportRemoteError(masterConnection, queryResult);
ReportRemoteError(masterConnection, queryResult, false);
ereport(ERROR, (errmsg("could not create a new empty shard on the remote node")));
}

View File

@ -141,7 +141,7 @@ MultiClientConnect(const char *nodeName, uint32 nodePort, const char *nodeDataba
}
else
{
ReportRemoteError(connection, NULL);
ReportRemoteError(connection, NULL, false);
PQfinish(connection);
connectionId = INVALID_CONNECTION_ID;
@ -194,7 +194,7 @@ MultiClientConnectStart(const char *nodeName, uint32 nodePort, const char *nodeD
}
else
{
ReportRemoteError(connection, NULL);
ReportRemoteError(connection, NULL, false);
PQfinish(connection);
connectionId = INVALID_CONNECTION_ID;
@ -249,7 +249,7 @@ MultiClientConnectPoll(int32 connectionId)
}
else if (pollingStatus == PGRES_POLLING_FAILED)
{
ReportRemoteError(connection, NULL);
ReportRemoteError(connection, NULL, false);
connectStatus = CLIENT_CONNECTION_BAD;
}
@ -433,7 +433,7 @@ MultiClientQueryResult(int32 connectionId, void **queryResult, int *rowCount,
}
else
{
ReportRemoteError(connection, result);
ReportRemoteError(connection, result, false);
PQclear(result);
}
@ -500,7 +500,7 @@ MultiClientBatchResult(int32 connectionId, void **queryResult, int *rowCount,
}
else
{
ReportRemoteError(connection, result);
ReportRemoteError(connection, result, false);
PQclear(result);
queryStatus = CLIENT_BATCH_QUERY_FAILED;
}
@ -585,7 +585,7 @@ MultiClientQueryStatus(int32 connectionId)
copyResults = true;
}
ReportRemoteError(connection, result);
ReportRemoteError(connection, result, false);
}
/* clear the result object */
@ -675,7 +675,7 @@ MultiClientCopyData(int32 connectionId, int32 fileDescriptor)
{
copyStatus = CLIENT_COPY_FAILED;
ReportRemoteError(connection, result);
ReportRemoteError(connection, result, false);
}
PQclear(result);
@ -685,7 +685,7 @@ MultiClientCopyData(int32 connectionId, int32 fileDescriptor)
/* received an error */
copyStatus = CLIENT_COPY_FAILED;
ReportRemoteError(connection, NULL);
ReportRemoteError(connection, NULL, false);
}
/* if copy out completed, make sure we drain all results from libpq */

View File

@ -267,7 +267,19 @@ ExecuteDistributedModify(Task *task)
result = PQexec(connection, task->queryString);
if (PQresultStatus(result) != PGRES_COMMAND_OK)
{
ReportRemoteError(connection, result);
char *sqlStateString = PQresultErrorField(result, PG_DIAG_SQLSTATE);
int category = 0;
bool raiseError = false;
/*
* If the error code is in constraint violation class, we want to
* fail fast because we must get the same error from all shard
* placements.
*/
category = ERRCODE_TO_CATEGORY(ERRCODE_INTEGRITY_CONSTRAINT_VIOLATION);
raiseError = SqlStateMatchesCategory(sqlStateString, category);
ReportRemoteError(connection, result, raiseError);
PQclear(result);
failedPlacementList = lappend(failedPlacementList, taskPlacement);
@ -451,14 +463,14 @@ SendQueryInSingleRowMode(PGconn *connection, char *query)
querySent = PQsendQuery(connection, query);
if (querySent == 0)
{
ReportRemoteError(connection, NULL);
ReportRemoteError(connection, NULL, false);
return false;
}
singleRowMode = PQsetSingleRowMode(connection);
if (singleRowMode == 0)
{
ReportRemoteError(connection, NULL);
ReportRemoteError(connection, NULL, false);
return false;
}
@ -505,7 +517,7 @@ StoreQueryResult(PGconn *connection, TupleDesc tupleDescriptor,
resultStatus = PQresultStatus(result);
if ((resultStatus != PGRES_SINGLE_TUPLE) && (resultStatus != PGRES_TUPLES_OK))
{
ReportRemoteError(connection, result);
ReportRemoteError(connection, result, false);
PQclear(result);
return false;

View File

@ -371,14 +371,14 @@ SendQueryToPlacements(char *shardQueryString, ShardConnections *shardConnections
result = PQexec(connection, "BEGIN");
if (PQresultStatus(result) != PGRES_COMMAND_OK)
{
ReportRemoteError(connection, result);
ReportRemoteError(connection, result, false);
ereport(ERROR, (errmsg("could not send query to shard placement")));
}
result = PQexec(connection, shardQueryString);
if (PQresultStatus(result) != PGRES_COMMAND_OK)
{
ReportRemoteError(connection, result);
ReportRemoteError(connection, result, false);
ereport(ERROR, (errmsg("could not send query to shard placement")));
}

View File

@ -59,7 +59,7 @@ initialize_remote_temp_table(PG_FUNCTION_ARGS)
result = PQexec(connection, POPULATE_TEMP_TABLE);
if (PQresultStatus(result) != PGRES_COMMAND_OK)
{
ReportRemoteError(connection, result);
ReportRemoteError(connection, result, false);
}
PQclear(result);
@ -90,7 +90,7 @@ count_remote_temp_table_rows(PG_FUNCTION_ARGS)
result = PQexec(connection, COUNT_TEMP_TABLE);
if (PQresultStatus(result) != PGRES_TUPLES_OK)
{
ReportRemoteError(connection, result);
ReportRemoteError(connection, result, false);
}
else
{

View File

@ -190,12 +190,36 @@ PurgeConnection(PGconn *connection)
}
/*
* SqlStateMatchesCategory returns true if the given sql state is in the given
* error category. Note that we use ERRCODE_TO_CATEGORY macro to determine error
* category of the sql state and expect the caller to use the same macro for the
* error category.
*/
bool
SqlStateMatchesCategory(char *sqlStateString, int category)
{
bool sqlStateMatchesCategory = false;
int sqlState = MAKE_SQLSTATE(sqlStateString[0], sqlStateString[1], sqlStateString[2],
sqlStateString[3], sqlStateString[4]);
int sqlStateCategory = ERRCODE_TO_CATEGORY(sqlState);
if (sqlStateCategory == category)
{
sqlStateMatchesCategory = true;
}
return sqlStateMatchesCategory;
}
/*
* ReportRemoteError retrieves various error fields from the a remote result and
* produces an error report at the WARNING level.
* produces an error report at the WARNING level or at the ERROR level if raise
* error is set.
*/
void
ReportRemoteError(PGconn *connection, PGresult *result)
ReportRemoteError(PGconn *connection, PGresult *result, bool raiseError)
{
char *sqlStateString = PQresultErrorField(result, PG_DIAG_SQLSTATE);
char *messagePrimary = PQresultErrorField(result, PG_DIAG_MESSAGE_PRIMARY);
@ -206,6 +230,7 @@ ReportRemoteError(PGconn *connection, PGresult *result)
char *nodeName = ConnectionGetOptionValue(connection, "host");
char *nodePort = ConnectionGetOptionValue(connection, "port");
int sqlState = ERRCODE_CONNECTION_FAILURE;
int errorLevel = WARNING;
if (sqlStateString != NULL)
{
@ -213,6 +238,11 @@ ReportRemoteError(PGconn *connection, PGresult *result)
sqlStateString[3], sqlStateString[4]);
}
if (raiseError)
{
errorLevel = ERROR;
}
/*
* If the PGresult did not contain a message, the connection may provide a
* suitable top level one. At worst, this is an empty string.
@ -233,18 +263,18 @@ ReportRemoteError(PGconn *connection, PGresult *result)
if (sqlState == ERRCODE_CONNECTION_FAILURE)
{
ereport(WARNING, (errcode(sqlState),
errmsg("connection failed to %s:%s", nodeName, nodePort),
errdetail("%s", messagePrimary)));
ereport(errorLevel, (errcode(sqlState),
errmsg("connection failed to %s:%s", nodeName, nodePort),
errdetail("%s", messagePrimary)));
}
else
{
ereport(WARNING, (errcode(sqlState), errmsg("%s", messagePrimary),
messageDetail ? errdetail("%s", messageDetail) : 0,
messageHint ? errhint("%s", messageHint) : 0,
messageContext ? errcontext("%s", messageContext) : 0,
errcontext("Error occurred on remote connection to %s:%s.",
nodeName, nodePort)));
ereport(errorLevel, (errcode(sqlState), errmsg("%s", messagePrimary),
messageDetail ? errdetail("%s", messageDetail) : 0,
messageHint ? errhint("%s", messageHint) : 0,
messageContext ? errcontext("%s", messageContext) : 0,
errcontext("Error occurred on remote connection to %s:%s.",
nodeName, nodePort)));
}
}
@ -316,7 +346,7 @@ ConnectToNode(char *nodeName, int32 nodePort, char *nodeUser)
/* warn if still erroring on final attempt */
if (attemptIndex == MAX_CONNECT_ATTEMPTS - 1)
{
ReportRemoteError(connection, NULL);
ReportRemoteError(connection, NULL, false);
}
PQfinish(connection);

View File

@ -75,7 +75,7 @@ PrepareRemoteTransactions(List *connectionList)
/* a failure to prepare is an implicit rollback */
transactionConnection->transactionState = TRANSACTION_STATE_CLOSED;
ReportRemoteError(connection, result);
ReportRemoteError(connection, result, false);
PQclear(result);
ereport(ERROR, (errcode(ERRCODE_IO_ERROR),

View File

@ -56,7 +56,8 @@ typedef struct NodeConnectionEntry
/* function declarations for obtaining and using a connection */
extern PGconn * GetOrEstablishConnection(char *nodeName, int32 nodePort);
extern void PurgeConnection(PGconn *connection);
extern void ReportRemoteError(PGconn *connection, PGresult *result);
extern bool SqlStateMatchesCategory(char *sqlStateString, int category);
extern void ReportRemoteError(PGconn *connection, PGresult *result, bool raiseError);
extern PGconn * ConnectToNode(char *nodeName, int nodePort, char *nodeUser);
extern char * ConnectionGetOptionValue(PGconn *connection, char *optionKeyword);
#endif /* CONNECTION_CACHE_H */

View File

@ -68,7 +68,7 @@ INSERT INTO pg_temp.:"proxy_tablename" (id) VALUES (1);
-- test copy with bad row in middle
\set VERBOSITY terse
COPY pg_temp.:"proxy_tablename" FROM stdin;
ERROR: could not modify any active placements
ERROR: null value in column "data" violates not-null constraint
\set VERBOSITY default
-- verify rows were copied to distributed table
SELECT * FROM insert_target ORDER BY id ASC;

View File

@ -149,10 +149,14 @@ ERROR: cannot plan INSERT using row with NULL value in partition column
-- INSERT violating column constraint
INSERT INTO limit_orders VALUES (18811, 'BUD', 14962, '2014-04-05 08:32:16', 'sell',
-5.00);
ERROR: could not modify any active placements
ERROR: new row for relation "limit_orders_750000" violates check constraint "limit_orders_limit_price_check"
DETAIL: Failing row contains (18811, BUD, 14962, 2014-04-05 08:32:16, sell, -5.00).
CONTEXT: Error occurred on remote connection to localhost:57637.
-- INSERT violating primary key constraint
INSERT INTO limit_orders VALUES (32743, 'LUV', 5994, '2001-04-16 03:37:28', 'buy', 0.58);
ERROR: could not modify any active placements
ERROR: duplicate key value violates unique constraint "limit_orders_pkey_750001"
DETAIL: Key (id)=(32743) already exists.
CONTEXT: Error occurred on remote connection to localhost:57638.
SET client_min_messages TO DEFAULT;
-- commands with non-constant partition values are unsupported
INSERT INTO limit_orders VALUES (random() * 100, 'ORCL', 152, '2011-08-25 11:50:45',
@ -261,26 +265,25 @@ SELECT kind, limit_price FROM limit_orders WHERE id = 246;
buy | 0.00
(1 row)
-- Test that shards which miss a modification are marked unhealthy
-- First: Mark all placements for a node as inactive
UPDATE pg_dist_shard_placement
SET shardstate = 3
WHERE nodename = 'localhost' AND
nodeport = :worker_1_port;
-- Second: Perform an INSERT to the remaining node
-- Test that on unique contraint violations, we fail fast
INSERT INTO limit_orders VALUES (275, 'ADR', 140, '2007-07-02 16:32:15', 'sell', 43.67);
-- Third: Mark the original placements as healthy again
UPDATE pg_dist_shard_placement
SET shardstate = 1
WHERE nodename = 'localhost' AND
nodeport = :worker_1_port;
-- Fourth: Perform the same INSERT (primary key violation)
INSERT INTO limit_orders VALUES (275, 'ADR', 140, '2007-07-02 16:32:15', 'sell', 43.67);
WARNING: duplicate key value violates unique constraint "limit_orders_pkey_750001"
ERROR: duplicate key value violates unique constraint "limit_orders_pkey_750001"
DETAIL: Key (id)=(275) already exists.
CONTEXT: Error occurred on remote connection to localhost:57638.
-- Last: Verify the insert worked but the placement with the PK violation is now unhealthy
SELECT count(*) FROM limit_orders WHERE id = 275;
-- Test that shards which miss a modification are marked unhealthy
-- First: Connect to the second worker node
\c - - - :worker_2_port
-- Second: Drop limit_orders shard on the second worker node
DROP TABLE limit_orders_750000;
-- Third: Connect back to master node
\c - - - :master_port
-- Fourth: Perform an INSERT on the remaining node
INSERT INTO limit_orders VALUES (276, 'ADR', 140, '2007-07-02 16:32:15', 'sell', 43.67);
WARNING: relation "limit_orders_750000" does not exist
CONTEXT: Error occurred on remote connection to localhost:57638.
-- Last: Verify the insert worked but the deleted placement is now unhealthy
SELECT count(*) FROM limit_orders WHERE id = 276;
count
-------
1

View File

@ -190,28 +190,26 @@ SELECT bidder_id FROM limit_orders WHERE id = 246;
UPDATE limit_orders SET (kind, limit_price) = ('buy', DEFAULT) WHERE id = 246;
SELECT kind, limit_price FROM limit_orders WHERE id = 246;
-- Test that on unique contraint violations, we fail fast
INSERT INTO limit_orders VALUES (275, 'ADR', 140, '2007-07-02 16:32:15', 'sell', 43.67);
INSERT INTO limit_orders VALUES (275, 'ADR', 140, '2007-07-02 16:32:15', 'sell', 43.67);
-- Test that shards which miss a modification are marked unhealthy
-- First: Mark all placements for a node as inactive
UPDATE pg_dist_shard_placement
SET shardstate = 3
WHERE nodename = 'localhost' AND
nodeport = :worker_1_port;
-- First: Connect to the second worker node
\c - - - :worker_2_port
-- Second: Perform an INSERT to the remaining node
INSERT INTO limit_orders VALUES (275, 'ADR', 140, '2007-07-02 16:32:15', 'sell', 43.67);
-- Second: Drop limit_orders shard on the second worker node
DROP TABLE limit_orders_750000;
-- Third: Mark the original placements as healthy again
UPDATE pg_dist_shard_placement
SET shardstate = 1
WHERE nodename = 'localhost' AND
nodeport = :worker_1_port;
-- Third: Connect back to master node
\c - - - :master_port
-- Fourth: Perform the same INSERT (primary key violation)
INSERT INTO limit_orders VALUES (275, 'ADR', 140, '2007-07-02 16:32:15', 'sell', 43.67);
-- Fourth: Perform an INSERT on the remaining node
INSERT INTO limit_orders VALUES (276, 'ADR', 140, '2007-07-02 16:32:15', 'sell', 43.67);
-- Last: Verify the insert worked but the placement with the PK violation is now unhealthy
SELECT count(*) FROM limit_orders WHERE id = 275;
-- Last: Verify the insert worked but the deleted placement is now unhealthy
SELECT count(*) FROM limit_orders WHERE id = 276;
SELECT count(*)
FROM pg_dist_shard_placement AS sp,
pg_dist_shard AS s