external
/
citus
mirror of https://github.com/citusdata/citus.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #1595 from citusdata/improve_deadlock_detection 

Improve deadlock detection
pull/1598/head
Önder Kalacı 2017-08-18 13:28:21 +03:00 committed by GitHub
parent 2cc46f3a0c b82f886ad3
commit 96391bea15
1 changed files with 26 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
src/backend/distributed/transaction/distributed_deadlock_detection.c
View File

@ -101,12 +101,29 @@ check_distributed_deadlocks(PG_FUNCTION_ARGS)
bool bool
CheckForDistributedDeadlocks(void) CheckForDistributedDeadlocks(void)
{ {
WaitGraph *waitGraph = BuildGlobalWaitGraph(); WaitGraph *waitGraph = NULL;
HTAB *adjacencyLists = BuildAdjacencyListsForWaitGraph(waitGraph); HTAB *adjacencyLists = NULL;
HASH_SEQ_STATUS status; HASH_SEQ_STATUS status;
TransactionNode *transactionNode = NULL; TransactionNode *transactionNode = NULL;
int edgeCount = waitGraph->edgeCount; int edgeCount = 0;
int localGroupId = GetLocalGroupId(); int localGroupId = GetLocalGroupId();
List *workerNodeList = ActiveReadableNodeList();
/*
* We don't need to do any distributed deadlock checking if there
* are no worker nodes. This might even be problematic for a non-mx
* worker node which has the same group id with its master (i.e., 0),
* which may erroneously decide to kill the deadlocks happening on it.
*/
if (list_length(workerNodeList) == 0)
{
return false;
}
waitGraph = BuildGlobalWaitGraph();
adjacencyLists = BuildAdjacencyListsForWaitGraph(waitGraph);
edgeCount = waitGraph->edgeCount;
/* /*
* We iterate on transaction nodes and search for deadlocks where the * We iterate on transaction nodes and search for deadlocks where the
@ -135,8 +152,12 @@ CheckForDistributedDeadlocks(void)
TransactionNode *youngestTransaction = transactionNode; TransactionNode *youngestTransaction = transactionNode;
ListCell *participantTransactionCell = NULL; ListCell *participantTransactionCell = NULL;
/* there should be at least two transactions to get into a deadlock */ /*
Assert(list_length(deadlockPath) > 1); * There should generally be at least two transactions to get into a
* deadlock. However, in case Citus gets into a self-deadlock, we may
* find a deadlock with a single transaction.
*/
Assert(list_length(deadlockPath) >= 1);
LogDistributedDeadlockDebugMessage("Distributed deadlock found among the " LogDistributedDeadlockDebugMessage("Distributed deadlock found among the "
"following distributed transactions:"); "following distributed transactions:");