def fetch_splits_from_shards(config, uri, slave_ok):
"""Internal method to fetch splits from shareded db
:returns: The splits
"""
logging.warn("WARNING getting splits that connect directly to the backend mongods is risky and might not produce correct results")
connection = get_connection(uri)
configDB = connection["config"]
shardsColl = configDB["shards"]
shardSet = []
cur = shardsColl.find()
try:
for row in cur:
host = row.get('host')
slashIndex = host.find("/")
if slashIndex > 0:
host = host[slashIndex + 1:]
shardSet.append(host)
splits = []
for host in shardSet:
new_uri = get_new_URI(uri,host,slave_ok)
config['input_uri'] = new_uri
splits += calculate_unsharded_splits(config,slave_ok,new_uri)
#I think this is better than commented way
return splits
'''
def __init__(self,stream,params):
from mongo_util import get_connection,get_collection
config = {}
for key, value in params.__dict__.iteritems():
config[key] = value
self.uri = config.get('output_uri')
self.conn = get_connection(self.uri)
self.coll = get_collection(self.uri)
self.key_name = config.get('job_output_key','_id')
self.value_name = config.get('job_output_value')
def runTest(self):
#put 20000 objects in a database, call for a split by hand, then a split by the class
conn = get_connection(
"mongodb://ec2-23-20-75-24.compute-1.amazonaws.com:27020/test.twitter"
)
db = conn[config.get('db_name')]
coll = db[config.get('collection_name')]
#print db.command("collstats", coll.full_name)
#NOTE: need to run this code once to populate the database, after that comment it out
'''
for i in range(40000):
post = {"name" : i, "date": datetime.datetime.utcnow()}
coll.insert(post)
'''
#print coll.count()
command = bson.son.SON()
command['splitVector'] = coll.full_name
command['maxChunkSize'] = config.get('split_size')
command['force'] = False
command['keyPattern'] = {'_id': 1}
#SON([('splitVector', u'test.twitter'), ('maxChunkSize', 1), ('keyPattern', {'_id': 1}), ('force', False)])
results = db.command(command)
man_splits = results.get("splitKeys")
assert results.get(
'ok') == 1.0, 'split command did not return with 1.0 ok'
print results
print 'man_splits = ', len(man_splits)
assert man_splits, 'no splitKeys returned'
#now do it through MongoSplit
splits = calculate_splits(config)
assert splits, "MongoSplitter did not return the right splits"
logging.info("Calculated %s MongoInputSplits" % len(splits))
#assert len(man_splits) + 1 == len(splits) , "MongoSplitter returned a different number of splits than manual splits"
'''
def runTest(self):
# put 20000 objects in a database, call for a split by hand, then a split by the class
conn = get_connection("mongodb://ec2-23-20-75-24.compute-1.amazonaws.com:27020/test.twitter")
db = conn[config.get("db_name")]
coll = db[config.get("collection_name")]
# print db.command("collstats", coll.full_name)
# NOTE: need to run this code once to populate the database, after that comment it out
"""
for i in range(40000):
post = {"name" : i, "date": datetime.datetime.utcnow()}
coll.insert(post)
"""
# print coll.count()
command = bson.son.SON()
command["splitVector"] = coll.full_name
command["maxChunkSize"] = config.get("split_size")
command["force"] = False
command["keyPattern"] = {"_id": 1}
# SON([('splitVector', u'test.twitter'), ('maxChunkSize', 1), ('keyPattern', {'_id': 1}), ('force', False)])
results = db.command(command)
man_splits = results.get("splitKeys")
assert results.get("ok") == 1.0, "split command did not return with 1.0 ok"
print results
print "man_splits = ", len(man_splits)
assert man_splits, "no splitKeys returned"
# now do it through MongoSplit
splits = calculate_splits(config)
assert splits, "MongoSplitter did not return the right splits"
logging.info("Calculated %s MongoInputSplits" % len(splits))
# assert len(man_splits) + 1 == len(splits) , "MongoSplitter returned a different number of splits than manual splits"
"""
def init_databases(config_file):
global userdata_db, plog_db, report_db
userdata_db = mongo_util.get_db('entities_main', config_file)
plog_db = mongo_util.get_connection('datastore', config_file)['kadb_pl']
report_db = mongo_util.get_db('reporting', config_file)
def fetch_splits_via_chunks(config, uri, useShards, slaveOk):
"""Retrieves split objects based on chunks in mongo
:returns: The splits
"""
originalQuery = config.get("query")
if useShards:
logging.warn("WARNING getting splits that connect directly to the \
backend mongods is risky and might not produce correct results")
logging.debug("fetch_splits_via_chunks: originalQuery: %s" % originalQuery)
connection = get_connection(uri)
configDB = connection["config"]
shardMap = {}
if useShards:
shardsColl = configDB["shards"]
cur = shardsColl.find()
try:
for row in cur:
host = row.get('host')
slashIndex = host.find("/")
if slashIndex > 0:
host = host[slashIndex + 1:]
shardMap[row.get('_id')] = host
logging.debug("MongoInputFormat.getSplitsUsingChunks(): shard map is: %s" % shardMap)
chunksCollection = configDB["chunks"]
logging.info(configDB.collection_names())
query = bson.son.SON()
uri_info = uri_parser.parse_uri(uri)
query["ns"] = uri_info['database'] + '.' + uri_info['collection']
cur = chunksCollection.find(query)
logging.info("query is ", query)
logging.info(cur.count())
logging.info(chunksCollection.find().count())
try:
numChunks = 0
splits = []
for row in cur:
numChunks += 1
minObj = row.get('min')
shardKeyQuery = bson.son.SON()
min = bson.son.SON()
max = bson.son.SON()
for key in minObj:
tMin = minObj[key]
tMax = (row.get('max'))[key]
#@to-do do type comparison first?
min[key] = tMin
max[key] = tMax
if originalQuery == None:
originalQuery = bson.son.SON()
shardKeyQuery["$query"] = originalQuery
shardKeyQuery["$min"] = min
shardKeyQuery["$max"] = max
inputURI = config.get("input_uri")
if useShards:
shardName = row.get('shard')
host = shardMap[shardName]
inputURI = get_new_URI(inputURI, host, slaveOk)
splits.append(MongoInputSplit(
inputURI,
config.get("input_key"),
shardKeyQuery,
config.get("fields"),
config.get("sort"),
config.get("limit", 0),
config.get("skip", 0),
config.get("timeout", True)))
# return splits in uri format for disco
return [s.format_uri_with_query() for s in splits]
