I am very puzzled as to the behavior of some multiprocessing code that is using psycopg2 to make queries in parallel to a postgres db.
Essentially, I am making the same query (with different params) to various partitions of a larger table. I am using multiprocessing.Pool to fork off a separate query.
My multiprocessing call looks like this:
pool = Pool(processes=num_procs)
results=pool.map(run_sql, params_list)
My run_sql code looks like this:
def run_sql(zip2):
conn = get_connection()
curs = conn.cursor()
print "conn: %s curs:%s pid=%s" % (id(conn), id(curs), os.getpid())
...
curs.execute(qry)
records = curs.fetchall()
def get_connection()
...
conn = psycopg2.connect(user=db_user, host=db_host,
dbname=db_name, password=db_pwd)
return conn
So my expectation is that each process would get a separate db connection via the call to get_connection() and that print id(conn) would display a distinct value. However, that doesn't seem to be the case and I am at a loss to explain it. Even print id(curs) is the same. Only print os.getpid() shows a difference. Does it somehow use the same connection for each forked process ?
conn: 4614554592 curs:4605160432 pid=46802
conn: 4614554592 curs:4605160432 pid=46808
conn: 4614554592 curs:4605160432 pid=46810
conn: 4614554592 curs:4605160432 pid=46784
conn: 4614554592 curs:4605160432 pid=46811
I think I've figured this out. The answer lies in the fact that multiprocessing in Python is shared-nothing so the entire memory space is copied, functions and all. Hence for each process, even though the pid is different, the memory spaces are copies of each other and the address of the connection within the memory space ends up being the same. The same reason is why declaring a global connection pool as I did initially was useless, each process ended up with its own connection pool with just 1 connection active at a time.
Related
I have a function that queries a large table for the purposes of indexing it... It creates a server-side cursor named "all_accounts".
def get_all_accounts(self):
cursor = self.get_cursor('all_accounts')
cursor.execute("SELECT * FROM account_summary LIMIT 20000;")
I then process those 2,000 or so at a time to insert into a NoSQL solution:
def index_docs(self, cursor):
while True:
# consume result over a series of iterations
# with each iteration fetching 2000 records
record_count = cursor.rowcount
records = cursor.fetchmany(size=2000)
if not records:
break
for r in records:
# do stuff
I'd like the index_docs function to be consuming the cursor fetchmany() calls in parallel x10 as my bottleneck is not caused by the target system, but rather the single threaded nature of my script. I have done a few async/worker things in the past, but the psycopg2 cursor seemed like it might be an issue. Thoughts?
I think you'll be safe if a single process/thread accesses the cursor and dishes out work to multiple worker processes that push to the other database. (At a quick glance, server-side cursors can't be shared between connections, but I could be wrong there.)
That is, something like this. Generally you'd use imap_unordered to iterate over a collection of single items (and use a higher chunksize than the default 1), but I think we can just as well use the batches here...
import multiprocessing
def get_batches(conn):
cursor = conn.get_cursor('all_accounts')
cursor.execute("SELECT * FROM account_summary LIMIT 20000;")
while True:
records = cursor.fetchmany(size=500)
if not records:
break
yield list(records)
def process_batch(batch):
# (this function is run in child processes)
for r in batch:
# ...
return "some arbitrary result"
def main():
conn = connect...()
with multiprocessing.Pool() as p:
batch_generator = get_batches(conn)
for result in p.imap_unordered(process_batch, get_batches):
print(result) # doesn't really matter
I followed the below code in order to implement a parallel select query on a postgres database:
https://tech.geoblink.com/2017/07/06/parallelizing-queries-in-postgresql-with-python/
My basic problem is that I have ~6k queries that need to be executed, and I am trying to optimise the execution of these select queries. Initially it was a single query with the where id in (...) contained all 6k predicate IDs but I ran into issues with the query using up > 4GB of RAM on the machine it ran on, so I decided to split it out into 6k individual queries which when synchronously keeps a steady memory usage. However it takes a lot longer to run time wise, which is less of an issue for my use case. Even so I am trying to reduce the time as much as possible.
This is what my code looks like:
class PostgresConnector(object):
def __init__(self, db_url):
self.db_url = db_url
self.engine = self.init_connection()
self.pool = self.init_pool()
def init_pool(self):
CPUS = multiprocessing.cpu_count()
return multiprocessing.Pool(CPUS)
def init_connection(self):
LOGGER.info('Creating Postgres engine')
return create_engine(self.db_url)
def run_parallel_queries(self, queries):
results = []
try:
for i in self.pool.imap_unordered(self.execute_parallel_query, queries):
results.append(i)
except Exception as exception:
LOGGER.error('Error whilst executing %s queries in parallel: %s', len(queries), exception)
raise
finally:
self.pool.close()
self.pool.join()
LOGGER.info('Parallel query ran producing %s sets of results of type: %s', len(results), type(results))
return list(chain.from_iterable(results))
def execute_parallel_query(self, query):
con = psycopg2.connect(self.db_url)
cur = con.cursor()
cur.execute(query)
records = cur.fetchall()
con.close()
return list(records)
However whenever this runs, I get the following error:
TypeError: can't pickle _thread.RLock objects
I've read lots of similar questions regarding the use of multiprocessing and pickleable objects but I cant for the life of me figure out what I am doing wrong.
The pool is generally one per process (which I believe is the best practise) but shared per instance of the connector class so that its not creating a pool for each use of the parallel_query method.
The top answer to a similar question:
Accessing a MySQL connection pool from Python multiprocessing
Shows an almost identical implementation to my own, except using MySql instead of Postgres.
Am I doing something wrong?
Thanks!
EDIT:
I've found this answer:
Python Postgres psycopg2 ThreadedConnectionPool exhausted
which is incredibly detailed and looks as though I have misunderstood what multiprocessing.Pool vs a connection pool such as ThreadedConnectionPool gives me. However in the first link it doesn't mention needing any connection pools etc. This solution seems good but seems A LOT of code for what I think is a fairly simple problem?
EDIT 2:
So the above link solves another problem, which I would have likely run into anyway so I'm glad I found that, but it doesnt solve the initial issue of not being able to use imap_unordered down to the pickling error. Very frustrating.
Lastly, I think its probably worth noting that this runs in Heroku, on a worker dyno, using Redis rq for scheduling, background tasks etc and a hosted instance of Postgres as the database.
To put it simply, postgres connection and sqlalchemy connection pool is thread safe, however they are not fork-safe.
If you want to use multiprocessing, you should initialize the engine in each child processes after the fork.
You should use multithreading instead if you want to share engines.
Refer to Thread and process safety in psycopg2 documentation:
libpq connections
shouldn’t be used by a forked processes, so when using a module such
as multiprocessing or a forking web deploy method such as FastCGI make
sure to create the connections after the fork.
If you are using multiprocessing.Pool, there is a keyword argument initializer which can be used to run code once on each child process. Try this:
class PostgresConnector(object):
def __init__(self, db_url):
self.db_url = db_url
self.pool = self.init_pool()
def init_pool(self):
CPUS = multiprocessing.cpu_count()
return multiprocessing.Pool(CPUS, initializer=self.init_connection(self.db_url))
#classmethod
def init_connection(cls, db_url):
def _init_connection():
LOGGER.info('Creating Postgres engine')
cls.engine = create_engine(db_url)
return _init_connection
def run_parallel_queries(self, queries):
results = []
try:
for i in self.pool.imap_unordered(self.execute_parallel_query, queries):
results.append(i)
except Exception as exception:
LOGGER.error('Error whilst executing %s queries in parallel: %s', len(queries), exception)
raise
finally:
pass
#self.pool.close()
#self.pool.join()
LOGGER.info('Parallel query ran producing %s sets of results of type: %s', len(results), type(results))
return list(chain.from_iterable(results))
def execute_parallel_query(self, query):
with self.engine.connect() as conn:
with conn.begin():
result = conn.execute(query)
return result.fetchall()
def __getstate__(self):
# this is a hack, if you want to remove this method, you should
# remove self.pool and just pass pool explicitly
self_dict = self.__dict__.copy()
del self_dict['pool']
return self_dict
Now, to address the XY problem.
Initially it was a single query with the where id in (...) contained
all 6k predicate IDs but I ran into issues with the query using up >
4GB of RAM on the machine it ran on, so I decided to split it out into
6k individual queries which when synchronously keeps a steady memory
usage.
What you may want to do instead is one of these options:
write a subquery that generates all 6000 IDs and use the subquery in your original bulk query.
as above, but write the subquery as a CTE
if your ID list comes from an external source (i.e. not from the database), then you can create a temporary table containing the 6000 IDs and then run your original bulk query against the temporary table
However, if you insist on running 6000 IDs through python, then the fastest query is likely neither to do all 6000 IDs in one go (which will run out of memory) nor to run 6000 individual queries. Instead, you may want to try to chunk the queries. Send 500 IDs at once for example. You will have to experiment with the chunk size to determine the largest number of IDs you can send at one time while still comfortably within your memory budget.
I'm trying to insert to update really big values of data in a MySQL db and in the same try, I was trying to see in the process list what is doing!
So I made the following script:
I have a modified db MySQL that takes care to connect. Everything is working fine unless I use multiprocesses, if I use multiprocessing I got an error at some time with "Lost connection to database".
The script is like:
from mysql import DB
import multiprocessing
def check writing(db):
result = db.execute("show full processlist").fethcall()
for i in result:
if i['State'] == "updating":
print i['Info']
def main(db):
# some work to create a big list of tuple called tuple
sql = "update `table_name` set `field` = %s where `primary_key_id` = %s"
monitor = multiprocessing.Process(target=check_writing,args=(db,)) # I create the monitor process
monitor.start()
db.execute_many(sql,tuple) # I start to modify table
monitor.terminate()
monitor.join
if __name__ == "__main__"
db = DB(host,user,password,database_name) # this way I create the object connected
main(db)
db.close()
And the a part of my mysql class is:
class DB:
def __init__(self,host,user,password,db_name)
self.db = MySQLdb.connect(host=host.... etc
def execute_many(self,sql,data):
c = self.db.cursor()
c.executemany(sql, data)
c.close()
self.db.commit()
As I said before, if I don't try to execute in check_writing, the script is working fine!
Maybe someone can explain me what is the cause and how can overcome? Also, I have problems trying to threadPool writing in MySQL using map (or map_async).
Do I miss something related to mysql?
There is a better way to approach that:
Connector/Python Connection Pooling:
mysql.connector.pooling module implements pooling.
A pool opens a number of connections and handles thread safety when providing connections to requesters.
The size of a connection pool is configurable at pool creation time. It cannot be resized thereafter.
it is possible to have multiple connection pools. This enables applications to support pools of connections to different MySQL servers, for example.
Check documentation here
I think your parallel processes are exhausting your mysql connections.
I use python multiprocessing processes to establish multiple connections to a postgreSQL database via psycopg.
Every process establishes a connection, creates a cursor, fetches an object from a mp.Queue and does some work on the database. If everything works fine, the changes are commited and the connection is closed.
If one of the processes however creates an error (e.g. an ADD COLUMN request fails, because the COLUMN is already present), all the processes seem to stop working.
import psycopg2
import multiprocessing as mp
import Queue
def connect():
C = psycopg2.connect(host = "myhost", user = "myuser", password = "supersafe", port = 62013, database = "db")
cur = C.cursor()
return C,cur
def commit_and_close(C,cur):
C.commit()
cur.close()
C.close()
def commit(C):
C.commit()
def sub(queue):
C,cur = connect()
while not queue.empty():
work_element = queue.get(timeout=1)
#do something with the work element, that might produce an SQL error
commit_and_close(C,cur)
return 0
if __name__ == '__main__':
job_queue = mp.Queue()
#Fill Job_queue
print 'Run'
for i in range(20):
p=mp.Process(target=sub, args=(job_queue))
p.start()
I can see, that processes are still alive (because the job_queue is still full), but no Network traffic / SQL actions are happening. Is it possible, that an SQL error blocks communication from other subprocesses? How can I prevent that happening?
As chance would have it, I was doing something similar today.
It shouldn't be that the state of one connection can affect a different one, so I don't think we should start there.
There is clearly a race condition in your queue handling. You check if the queue is empty and then try to get a statement to execute. With multiple readers one of the others could empty the queue leaving the others all blocking on their queue.get. If the queue is empty when they all lock up then I would suspect this.
You also never join your processes back when they complete. I'm not sure what effect that would have in the larger picture, but it's probably good practice to clean up.
The other thing that might be happening is that your error-ing process is not rolling back properly. That might leave other transactions waiting to see if it completes or rolls back. They can wait for quite a long time by default but you can configure it.
To see what is happening, fire up psql and check out two useful system views pg_stat_activity and pg_locks. That should show where the cause lies.
The SQLite documentation says (here) that you can avoid checkpoint pauses in WAL-mode by running the checkpoints on a separate thread. I tried this, and it doesn't appear to work: the '-wal' file grows without bound, it is unclear whether anything is actually getting copied back into the main database file, and (most important) after the -wal file has gotten big enough (over a gigabyte) the main thread starts having to wait for the checkpointer.
In my application the main thread continuously does something essentially equivalent to this, where generate_data is going to spit out order of a million rows to be inserted:
db = sqlite3.connect("database.db")
cursor = db.cursor()
cursor.execute("PRAGMA wal_autocheckpoint = 0")
for datum in generate_data():
# It is a damned shame that there is no way to do this in one operation.
cursor.execute("SELECT id FROM strings WHERE str = ?", (datum.text,))
row = cursor.fetchone()
if row is not None:
id = row[0]
else:
cur.execute("INSERT INTO strings VALUES(NULL, ?)", (datum.text,))
id = cur.lastrowid
cursor.execute("INSERT INTO data VALUES (?, ?, ?)",
(id, datum.foo, datum.bar))
batch_size += 1
if batch_size > batch_limit:
db.commit()
batch_size = 0
and the checkpoint thread does this:
db = sqlite3.connect("database.db")
cursor = db.cursor()
cursor.execute("PRAGMA wal_autocheckpoint = 0")
while True:
time.sleep(10)
cursor.execute("PRAGMA wal_checkpoint(PASSIVE)")
(Being on separate threads, they have to have separate connections to the database, because pysqlite doesn't support sharing a connection among multiple threads.) Changing to a FULL or RESTART checkpoint does not help - then the checkpoints just fail.
How do I make this actually work? Desiderata are: 1) main thread never has to wait, 2) journal file does not grow without bound.
Checkpointing needs to lock the entire database, so all other readers and writes would have to be blocked.
(A passive checkpoint just aborts.)
So running checkpointing in a separate thread does not increase concurrency.
(The SQLite documentation suggests this only because the main thread might no be designed to handle checkpointing at idle moments.)
If you continuously access the database, you cannot checkpoint.
If your batch operations make the WAL file grow too big, you should insert explicit checkpoints into that loop (or rely on autocheckpointing).