Cache worker node array for faster iteration

pull/1996/head
Marco Slot 2018-02-04 12:04:36 +01:00
parent 65fca44f4f
commit 40d715d494
1 changed files with 43 additions and 11 deletions

View File

@ -146,6 +146,8 @@ static HTAB *DistShardCacheHash = NULL;
/* Hash table for informations about worker nodes */
static HTAB *WorkerNodeHash = NULL;
static WorkerNode **WorkerNodeArray = NULL;
static int WorkerNodeCount = 0;
static bool workerNodeHashValid = false;
/* default value is -1, for coordinator it's 0 and for worker nodes > 0 */
@ -168,6 +170,7 @@ static ShardInterval ** SortShardIntervalArray(ShardInterval **shardIntervalArra
shardIntervalSortCompareFunction);
static bool HasUniformHashDistribution(ShardInterval **shardIntervalArray,
int shardIntervalArrayLength);
static void PrepareWorkerNodeCache(void);
static bool HasUninitializedShardInterval(ShardInterval **sortedShardIntervalArray,
int shardCount);
static bool CheckInstalledVersion(int elevel);
@ -511,15 +514,14 @@ ResolveGroupShardPlacement(GroupShardPlacement *groupShardPlacement,
static WorkerNode *
LookupNodeForGroup(uint32 groupId)
{
WorkerNode *workerNode = NULL;
HASH_SEQ_STATUS status;
HTAB *workerNodeHash = GetWorkerNodeHash();
bool foundAnyNodes = false;
int workerNodeIndex = 0;
hash_seq_init(&status, workerNodeHash);
PrepareWorkerNodeCache();
while ((workerNode = hash_seq_search(&status)) != NULL)
for (workerNodeIndex = 0; workerNodeIndex < WorkerNodeCount; workerNodeIndex++)
{
WorkerNode *workerNode = WorkerNodeArray[workerNodeIndex];
uint32 workerNodeGroupId = workerNode->groupId;
if (workerNodeGroupId != groupId)
{
@ -530,7 +532,6 @@ LookupNodeForGroup(uint32 groupId)
if (WorkerNodeIsReadable(workerNode))
{
hash_seq_term(&status);
return workerNode;
}
}
@ -2463,12 +2464,27 @@ InitializeDistTableCache(void)
/*
* GetWorkerNodeHash is a wrapper around InitializeWorkerNodeCache(). It
* triggers InitializeWorkerNodeCache when the workerHash is invalid. Otherwise,
* it returns the hash.
* GetWorkerNodeHash returns the worker node data as a hash with the nodename and
* nodeport as a key.
*
* The hash is returned from the cache, if the cache is not (yet) valid, it is first
* rebuilt.
*/
HTAB *
GetWorkerNodeHash(void)
{
PrepareWorkerNodeCache();
return WorkerNodeHash;
}
/*
* PrepareWorkerNodeCache makes sure the worker node data from pg_dist_node is cached,
* if it is not already cached.
*/
static void
PrepareWorkerNodeCache(void)
{
InitializeCaches(); /* ensure relevant callbacks are registered */
@ -2490,8 +2506,6 @@ GetWorkerNodeHash(void)
workerNodeHashValid = true;
}
return WorkerNodeHash;
}
@ -2510,6 +2524,9 @@ InitializeWorkerNodeCache(void)
int hashFlags = 0;
long maxTableSize = (long) MaxWorkerNodesTracked;
bool includeNodesFromOtherClusters = false;
int newWorkerNodeCount = 0;
WorkerNode **newWorkerNodeArray = NULL;
int workerNodeIndex = 0;
InitializeCaches();
@ -2534,6 +2551,10 @@ InitializeWorkerNodeCache(void)
/* read the list from pg_dist_node */
workerNodeList = ReadWorkerNodes(includeNodesFromOtherClusters);
newWorkerNodeCount = list_length(workerNodeList);
newWorkerNodeArray = MemoryContextAlloc(CacheMemoryContext,
sizeof(WorkerNode *) * newWorkerNodeCount);
/* iterate over the worker node list */
foreach(workerNodeCell, workerNodeList)
{
@ -2558,6 +2579,8 @@ InitializeWorkerNodeCache(void)
workerNode->nodeRole = currentNode->nodeRole;
strlcpy(workerNode->nodeCluster, currentNode->nodeCluster, NAMEDATALEN);
newWorkerNodeArray[workerNodeIndex++] = workerNode;
if (handleFound)
{
ereport(WARNING, (errmsg("multiple lines for worker node: \"%s:%u\"",
@ -2571,6 +2594,15 @@ InitializeWorkerNodeCache(void)
/* now, safe to destroy the old hash */
hash_destroy(oldWorkerNodeHash);
if (WorkerNodeArray != NULL)
{
pfree(WorkerNodeArray);
}
WorkerNodeCount = newWorkerNodeCount;
WorkerNodeArray = newWorkerNodeArray;
}