/*
* pg_get_serverdef_string finds the foreign server that corresponds to the
* given foreign tableId, and returns this server's definition.
*/
char *
pg_get_serverdef_string(Oid tableRelationId)
{
ForeignTable *foreignTable = GetForeignTable(tableRelationId);
ForeignServer *server = GetForeignServer(foreignTable->serverid);
ForeignDataWrapper *foreignDataWrapper = GetForeignDataWrapper(server->fdwid);
StringInfoData buffer = { NULL, 0, 0, 0 };
initStringInfo(&buffer);
appendStringInfo(&buffer, "CREATE SERVER %s", quote_identifier(server->servername));
if (server->servertype != NULL)
{
appendStringInfo(&buffer, " TYPE %s",
quote_literal_cstr(server->servertype));
}
if (server->serverversion != NULL)
{
appendStringInfo(&buffer, " VERSION %s",
quote_literal_cstr(server->serverversion));
}
appendStringInfo(&buffer, " FOREIGN DATA WRAPPER %s",
quote_identifier(foreignDataWrapper->fdwname));
/* append server options, if any */
AppendOptionListToString(&buffer, server->options);
return (buffer.data);
}
/*
* AppendOptionListToString converts the option list to its textual format, and
* appends this text to the given string buffer.
*/
void
AppendOptionListToString(StringInfo stringBuffer, List *optionList)
{
if (optionList != NIL)
{
ListCell *optionCell = NULL;
bool firstOptionPrinted = false;
appendStringInfo(stringBuffer, " OPTIONS (");
foreach(optionCell, optionList)
{
DefElem *option = (DefElem *) lfirst(optionCell);
char *optionName = option->defname;
char *optionValue = defGetString(option);
if (firstOptionPrinted)
{
appendStringInfo(stringBuffer, ", ");
}
firstOptionPrinted = true;
appendStringInfo(stringBuffer, "%s ", quote_identifier(optionName));
appendStringInfo(stringBuffer, "%s", quote_literal_cstr(optionValue));
}
/*
* GenerateAttachShardPartitionCommand generates command to attach a child table
* table to its parent in a partitioning hierarchy.
*/
char *
GenerateAttachShardPartitionCommand(ShardInterval *shardInterval)
{
Oid schemaId = get_rel_namespace(shardInterval->relationId);
char *schemaName = get_namespace_name(schemaId);
char *escapedSchemaName = quote_literal_cstr(schemaName);
char *command = GenerateAlterTableAttachPartitionCommand(shardInterval->relationId);
char *escapedCommand = quote_literal_cstr(command);
int shardIndex = ShardIndex(shardInterval);
Oid parentSchemaId = InvalidOid;
char *parentSchemaName = NULL;
char *escapedParentSchemaName = NULL;
uint64 parentShardId = INVALID_SHARD_ID;
StringInfo attachPartitionCommand = makeStringInfo();
Oid parentRelationId = PartitionParentOid(shardInterval->relationId);
if (parentRelationId == InvalidOid)
{
ereport(ERROR, (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("cannot attach partition"),
errdetail("Referenced relation cannot be found.")));
}
parentSchemaId = get_rel_namespace(parentRelationId);
parentSchemaName = get_namespace_name(parentSchemaId);
escapedParentSchemaName = quote_literal_cstr(parentSchemaName);
parentShardId = ColocatedShardIdInRelation(parentRelationId, shardIndex);
appendStringInfo(attachPartitionCommand,
WORKER_APPLY_INTER_SHARD_DDL_COMMAND, parentShardId,
escapedParentSchemaName, shardInterval->shardId,
escapedSchemaName, escapedCommand);
return attachPartitionCommand->data;
}
static PyObject *
PLy_quote_literal(PyObject *self, PyObject *args)
{
const char *str;
char *quoted;
PyObject *ret;
if (!PyArg_ParseTuple(args, "s:quote_literal", &str))
return NULL;
quoted = quote_literal_cstr(str);
ret = PyString_FromString(quoted);
pfree(quoted);
return ret;
}
Datum
pg_tablespace_size_name(PG_FUNCTION_ARGS)
{
Name tblspcName = PG_GETARG_NAME(0);
Oid tblspcOid = get_tablespace_oid(NameStr(*tblspcName), false);
int64 size;
size = calculate_tablespace_size(tblspcOid);
if (Gp_role == GP_ROLE_DISPATCH)
{
char *sql;
sql = psprintf("select pg_catalog.pg_tablespace_size(%s)",
quote_literal_cstr(NameStr(*tblspcName)));
size += get_size_from_segDBs(sql);
}
if (size < 0)
PG_RETURN_NULL();
PG_RETURN_INT64(size);
}
static Tuplestorestate *
build_tuplestore_recursively(char *key_fld,
char *parent_key_fld,
char *relname,
char *orderby_fld,
char *branch_delim,
char *start_with,
char *branch,
int level,
int *serial,
int max_depth,
bool show_branch,
bool show_serial,
MemoryContext per_query_ctx,
AttInMetadata *attinmeta,
Tuplestorestate *tupstore)
{
TupleDesc tupdesc = attinmeta->tupdesc;
int ret;
int proc;
int serial_column;
StringInfoData sql;
char **values;
char *current_key;
char *current_key_parent;
char current_level[INT32_STRLEN];
char serial_str[INT32_STRLEN];
char *current_branch;
HeapTuple tuple;
if (max_depth > 0 && level > max_depth)
return tupstore;
initStringInfo(&sql);
/* Build initial sql statement */
if (!show_serial)
{
appendStringInfo(&sql, "SELECT %s, %s FROM %s WHERE %s = %s AND %s IS NOT NULL AND %s <> %s",
key_fld,
parent_key_fld,
relname,
parent_key_fld,
quote_literal_cstr(start_with),
key_fld, key_fld, parent_key_fld);
serial_column = 0;
}
else
{
appendStringInfo(&sql, "SELECT %s, %s FROM %s WHERE %s = %s AND %s IS NOT NULL AND %s <> %s ORDER BY %s",
key_fld,
parent_key_fld,
relname,
parent_key_fld,
quote_literal_cstr(start_with),
key_fld, key_fld, parent_key_fld,
orderby_fld);
serial_column = 1;
}
if (show_branch)
values = (char **) palloc((CONNECTBY_NCOLS + serial_column) * sizeof(char *));
else
values = (char **) palloc((CONNECTBY_NCOLS_NOBRANCH + serial_column) * sizeof(char *));
/* First time through, do a little setup */
if (level == 0)
{
/* root value is the one we initially start with */
values[0] = start_with;
/* root value has no parent */
values[1] = NULL;
/* root level is 0 */
sprintf(current_level, "%d", level);
values[2] = current_level;
/* root branch is just starting root value */
if (show_branch)
values[3] = start_with;
/* root starts the serial with 1 */
if (show_serial)
{
sprintf(serial_str, "%d", (*serial)++);
if (show_branch)
values[4] = serial_str;
else
values[3] = serial_str;
}
/* construct the tuple */
tuple = BuildTupleFromCStrings(attinmeta, values);
/* now store it */
tuplestore_puttuple(tupstore, tuple);
/* increment level */
level++;
}
/* Retrieve the desired rows */
ret = SPI_execute(sql.data, true, 0);
proc = SPI_processed;
/* Check for qualifying tuples */
if ((ret == SPI_OK_SELECT) && (proc > 0))
{
HeapTuple spi_tuple;
SPITupleTable *tuptable = SPI_tuptable;
TupleDesc spi_tupdesc = tuptable->tupdesc;
int i;
StringInfoData branchstr;
StringInfoData chk_branchstr;
StringInfoData chk_current_key;
/* First time through, do a little more setup */
if (level == 0)
{
/*
* Check that return tupdesc is compatible with the one we got
* from the query, but only at level 0 -- no need to check more
* than once
*/
if (!compatConnectbyTupleDescs(tupdesc, spi_tupdesc))
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
errmsg("invalid return type"),
errdetail("Return and SQL tuple descriptions are " \
"incompatible.")));
}
initStringInfo(&branchstr);
initStringInfo(&chk_branchstr);
initStringInfo(&chk_current_key);
for (i = 0; i < proc; i++)
{
/* initialize branch for this pass */
appendStringInfo(&branchstr, "%s", branch);
appendStringInfo(&chk_branchstr, "%s%s%s", branch_delim, branch, branch_delim);
/* get the next sql result tuple */
spi_tuple = tuptable->vals[i];
/* get the current key and parent */
current_key = SPI_getvalue(spi_tuple, spi_tupdesc, 1);
appendStringInfo(&chk_current_key, "%s%s%s", branch_delim, current_key, branch_delim);
current_key_parent = pstrdup(SPI_getvalue(spi_tuple, spi_tupdesc, 2));
/* get the current level */
sprintf(current_level, "%d", level);
/* check to see if this key is also an ancestor */
if (strstr(chk_branchstr.data, chk_current_key.data))
elog(ERROR, "infinite recursion detected");
/* OK, extend the branch */
appendStringInfo(&branchstr, "%s%s", branch_delim, current_key);
current_branch = branchstr.data;
/* build a tuple */
values[0] = pstrdup(current_key);
values[1] = current_key_parent;
values[2] = current_level;
if (show_branch)
values[3] = current_branch;
if (show_serial)
{
sprintf(serial_str, "%d", (*serial)++);
if (show_branch)
values[4] = serial_str;
else
values[3] = serial_str;
}
tuple = BuildTupleFromCStrings(attinmeta, values);
xpfree(current_key);
xpfree(current_key_parent);
/* store the tuple for later use */
tuplestore_puttuple(tupstore, tuple);
heap_freetuple(tuple);
/* recurse using current_key_parent as the new start_with */
tupstore = build_tuplestore_recursively(key_fld,
parent_key_fld,
relname,
orderby_fld,
branch_delim,
values[0],
current_branch,
level + 1,
serial,
max_depth,
show_branch,
show_serial,
per_query_ctx,
attinmeta,
tupstore);
/* reset branch for next pass */
resetStringInfo(&branchstr);
resetStringInfo(&chk_branchstr);
resetStringInfo(&chk_current_key);
}
xpfree(branchstr.data);
xpfree(chk_branchstr.data);
xpfree(chk_current_key.data);
}
return tupstore;
}
/*
* master_append_table_to_shard appends the given table's contents to the given
* shard, and updates shard metadata on the master node. If the function fails
* to append table data to all shard placements, it doesn't update any metadata
* and errors out. Else if the function fails to append table data to some of
* the shard placements, it marks those placements as invalid. These invalid
* placements will get cleaned up during shard rebalancing.
*/
Datum
master_append_table_to_shard(PG_FUNCTION_ARGS)
{
uint64 shardId = PG_GETARG_INT64(0);
text *sourceTableNameText = PG_GETARG_TEXT_P(1);
text *sourceNodeNameText = PG_GETARG_TEXT_P(2);
uint32 sourceNodePort = PG_GETARG_UINT32(3);
char *sourceTableName = text_to_cstring(sourceTableNameText);
char *sourceNodeName = text_to_cstring(sourceNodeNameText);
char *shardName = NULL;
List *shardPlacementList = NIL;
List *succeededPlacementList = NIL;
List *failedPlacementList = NIL;
ListCell *shardPlacementCell = NULL;
ListCell *failedPlacementCell = NULL;
uint64 newShardSize = 0;
uint64 shardMaxSizeInBytes = 0;
float4 shardFillLevel = 0.0;
char partitionMethod = 0;
ShardInterval *shardInterval = LoadShardInterval(shardId);
Oid relationId = shardInterval->relationId;
bool cstoreTable = CStoreTable(relationId);
char storageType = shardInterval->storageType;
EnsureTablePermissions(relationId, ACL_INSERT);
if (storageType != SHARD_STORAGE_TABLE && !cstoreTable)
{
ereport(ERROR, (errmsg("cannot append to shardId " UINT64_FORMAT, shardId),
errdetail("The underlying shard is not a regular table")));
}
partitionMethod = PartitionMethod(relationId);
if (partitionMethod == DISTRIBUTE_BY_HASH)
{
ereport(ERROR, (errmsg("cannot append to shardId " UINT64_FORMAT, shardId),
errdetail("We currently don't support appending to shards "
"in hash-partitioned tables")));
}
/*
* We lock on the shardId, but do not unlock. When the function returns, and
* the transaction for this function commits, this lock will automatically
* be released. This ensures appends to a shard happen in a serial manner.
*/
LockShardResource(shardId, AccessExclusiveLock);
/* if shard doesn't have an alias, extend regular table name */
shardName = LoadShardAlias(relationId, shardId);
if (shardName == NULL)
{
shardName = get_rel_name(relationId);
AppendShardIdToName(&shardName, shardId);
}
shardPlacementList = FinalizedShardPlacementList(shardId);
if (shardPlacementList == NIL)
{
ereport(ERROR, (errmsg("could not find any shard placements for shardId "
UINT64_FORMAT, shardId),
errhint("Try running master_create_empty_shard() first")));
}
/* issue command to append table to each shard placement */
foreach(shardPlacementCell, shardPlacementList)
{
ShardPlacement *shardPlacement = (ShardPlacement *) lfirst(shardPlacementCell);
char *workerName = shardPlacement->nodeName;
uint32 workerPort = shardPlacement->nodePort;
List *queryResultList = NIL;
StringInfo workerAppendQuery = makeStringInfo();
appendStringInfo(workerAppendQuery, WORKER_APPEND_TABLE_TO_SHARD,
quote_literal_cstr(shardName),
quote_literal_cstr(sourceTableName),
quote_literal_cstr(sourceNodeName), sourceNodePort);
/* inserting data should be performed by the current user */
queryResultList = ExecuteRemoteQuery(workerName, workerPort, NULL,
workerAppendQuery);
if (queryResultList != NIL)
{
succeededPlacementList = lappend(succeededPlacementList, shardPlacement);
}
else
{
failedPlacementList = lappend(failedPlacementList, shardPlacement);
}
}