|
|
@ -19,6 +19,7 @@ import gevent.queue
|
|
|
|
from common import ensure_directory
|
|
|
|
from common import ensure_directory
|
|
|
|
|
|
|
|
|
|
|
|
from girc import Client
|
|
|
|
from girc import Client
|
|
|
|
|
|
|
|
from monotonic import monotonic
|
|
|
|
|
|
|
|
|
|
|
|
# How long each batch is
|
|
|
|
# How long each batch is
|
|
|
|
BATCH_INTERVAL = 60
|
|
|
|
BATCH_INTERVAL = 60
|
|
|
@ -239,25 +240,44 @@ def get_batch_files(path, batch_time):
|
|
|
|
]
|
|
|
|
]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def merge_all(path):
|
|
|
|
def merge_all(path, interval=None, stopping=None):
|
|
|
|
"""Repeatedly scans the batch directory for batch files with the same batch time, and merges them.
|
|
|
|
"""Repeatedly scans the batch directory for batch files with the same batch time, and merges them.
|
|
|
|
Returns once it finds no duplicate files."""
|
|
|
|
By default, returns once it finds no duplicate files.
|
|
|
|
while True:
|
|
|
|
If interval is given, re-scans after that number of seconds.
|
|
|
|
by_time = {}
|
|
|
|
If a gevent.event.Event() is passed in as stopping arg, returns when that event is set.
|
|
|
|
for hour in os.listdir(path):
|
|
|
|
"""
|
|
|
|
for name in os.listdir(os.path.join(path, hour)):
|
|
|
|
if stopping is None:
|
|
|
|
if not name.endswith(".json"):
|
|
|
|
# nothing will ever set this, but it's easier to not special-case it everywhere
|
|
|
|
continue
|
|
|
|
stopping = gevent.event.Event()
|
|
|
|
time = "-".join(name.split("-")[:3])
|
|
|
|
while not stopping.is_set():
|
|
|
|
timestamp = "{}:{}".format(hour, time)
|
|
|
|
start = monotonic()
|
|
|
|
by_time[timestamp] = by_time.get(timestamp, 0) + 1
|
|
|
|
# loop until no changes
|
|
|
|
if not any(count > 1 for timestamp, count in by_time.items()):
|
|
|
|
while True:
|
|
|
|
|
|
|
|
logging.debug("Scanning for merges")
|
|
|
|
|
|
|
|
by_time = {}
|
|
|
|
|
|
|
|
for hour in os.listdir(path):
|
|
|
|
|
|
|
|
for name in os.listdir(os.path.join(path, hour)):
|
|
|
|
|
|
|
|
if not name.endswith(".json"):
|
|
|
|
|
|
|
|
continue
|
|
|
|
|
|
|
|
time = "-".join(name.split("-")[:3])
|
|
|
|
|
|
|
|
timestamp = "{}:{}".format(hour, time)
|
|
|
|
|
|
|
|
by_time[timestamp] = by_time.get(timestamp, 0) + 1
|
|
|
|
|
|
|
|
if not any(count > 1 for timestamp, count in by_time.items()):
|
|
|
|
|
|
|
|
logging.info("All batches are merged")
|
|
|
|
|
|
|
|
break
|
|
|
|
|
|
|
|
for timestamp, count in by_time.items():
|
|
|
|
|
|
|
|
if count > 1:
|
|
|
|
|
|
|
|
logging.info("Merging {} batches at time {}".format(count, timestamp))
|
|
|
|
|
|
|
|
batch_time = timegm(time.strptime(timestamp, "%Y-%m-%dT%H:%M:%S"))
|
|
|
|
|
|
|
|
merge_batch_files(path, batch_time)
|
|
|
|
|
|
|
|
if interval is None:
|
|
|
|
|
|
|
|
# one-shot
|
|
|
|
break
|
|
|
|
break
|
|
|
|
for timestamp, count in by_time.items():
|
|
|
|
next_run = start + interval
|
|
|
|
if count > 1:
|
|
|
|
remaining = next_run - monotonic()
|
|
|
|
logging.info("Merging {} batches at time {}".format(count, timestamp))
|
|
|
|
if remaining > 0:
|
|
|
|
batch_time = timegm(time.strptime(timestamp, "%Y-%m-%dT%H:%M:%S"))
|
|
|
|
logging.debug("Waiting {}s for next merge scan".format(remaining))
|
|
|
|
merge_batch_files(path, batch_time)
|
|
|
|
stopping.wait(remaining)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def merge_batch_files(path, batch_time):
|
|
|
|
def merge_batch_files(path, batch_time):
|
|
|
@ -399,7 +419,7 @@ def merge_messages(left, right):
|
|
|
|
return result
|
|
|
|
return result
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def main(channel, nick, oauth_token_path, base_dir='/mnt', name=None):
|
|
|
|
def main(channel, nick, oauth_token_path, base_dir='/mnt', name=None, merge_interval=60):
|
|
|
|
with open(oauth_token_path) as f:
|
|
|
|
with open(oauth_token_path) as f:
|
|
|
|
oauth_token = f.read()
|
|
|
|
oauth_token = f.read()
|
|
|
|
# To ensure uniqueness even if multiple instances are running on the same host,
|
|
|
|
# To ensure uniqueness even if multiple instances are running on the same host,
|
|
|
@ -411,13 +431,19 @@ def main(channel, nick, oauth_token_path, base_dir='/mnt', name=None):
|
|
|
|
stopping = gevent.event.Event()
|
|
|
|
stopping = gevent.event.Event()
|
|
|
|
gevent.signal_handler(signal.SIGTERM, stopping.set)
|
|
|
|
gevent.signal_handler(signal.SIGTERM, stopping.set)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
merger = gevent.spawn(merge_all,
|
|
|
|
|
|
|
|
os.path.join(base_dir, channel, "chat"),
|
|
|
|
|
|
|
|
interval=merge_interval,
|
|
|
|
|
|
|
|
stopping=stopping
|
|
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
logging.info("Starting")
|
|
|
|
logging.info("Starting")
|
|
|
|
for index in count():
|
|
|
|
for index in count():
|
|
|
|
# To ensure uniqueness between clients, include a client number
|
|
|
|
# To ensure uniqueness between clients, include a client number
|
|
|
|
archiver = Archiver("{}.{}".format(name, index), base_dir, channel, nick, oauth_token)
|
|
|
|
archiver = Archiver("{}.{}".format(name, index), base_dir, channel, nick, oauth_token)
|
|
|
|
worker = gevent.spawn(archiver.run)
|
|
|
|
worker = gevent.spawn(archiver.run)
|
|
|
|
# wait for either graceful exit, error, or for a signal from the archiver that a reconnect was requested
|
|
|
|
# wait for either graceful exit, error, or for a signal from the archiver that a reconnect was requested
|
|
|
|
gevent.wait([stopping, worker, archiver.got_reconnect], count=1)
|
|
|
|
gevent.wait([stopping, worker, merger, archiver.got_reconnect], count=1)
|
|
|
|
if stopping.is_set():
|
|
|
|
if stopping.is_set():
|
|
|
|
archiver.stop()
|
|
|
|
archiver.stop()
|
|
|
|
worker.get()
|
|
|
|
worker.get()
|
|
|
@ -427,7 +453,13 @@ def main(channel, nick, oauth_token_path, base_dir='/mnt', name=None):
|
|
|
|
if archiver.got_reconnect.is_set():
|
|
|
|
if archiver.got_reconnect.is_set():
|
|
|
|
logging.info("Got RECONNECT, creating new client while waiting for old one to finish")
|
|
|
|
logging.info("Got RECONNECT, creating new client while waiting for old one to finish")
|
|
|
|
continue
|
|
|
|
continue
|
|
|
|
# the only remaining case is that the worker failed, re-raise
|
|
|
|
# the only remaining case is that something failed. stop everything and re-raise.
|
|
|
|
|
|
|
|
logging.warning("Stopping due to worker dying")
|
|
|
|
|
|
|
|
stopping.set()
|
|
|
|
|
|
|
|
worker.join()
|
|
|
|
|
|
|
|
merger.join()
|
|
|
|
|
|
|
|
# at least one of these two should raise
|
|
|
|
worker.get()
|
|
|
|
worker.get()
|
|
|
|
assert False, "Not stopping, but worker exited successfully"
|
|
|
|
merger.get()
|
|
|
|
|
|
|
|
assert False, "Worker unexpectedly exited successfully"
|
|
|
|
logging.info("Gracefully stopped")
|
|
|
|
logging.info("Gracefully stopped")
|
|
|
|