Closing Google Cloud PubSub client with future running - google-cloud-pubsub

I'm using Cloud Pubsub and for one of our systems I'm starting to get "Too many files open". lsof shows a tons of requests to Google Cloud, which I'm pretty sure are pubsub.
Googling led me to https://github.com/googleapis/google-cloud-python/issues/5523 which indicates I need to close the transport explicitly.
The problem is that I'm using a helper python package (which is invoked by about 50-100 other services) to publish my messages, which looks roughly like this:
def pubsub_callback(future):
message_id = future.result()
LOGGER.info("Successfully published %s", message_id)
def send_oneoff_pubsub_message(self, client=None):
if not client:
client = self.get_client('pubsubpub') # Creates a pubsub publisher client
future = client.publish({...})
try:
future.exception(timeout=10)
except Exception as exc:
print("error")
future.add_done_callback(pubsub_callback)
Now in many places we're slowly refactoring to explicitly create a client outside of the function (so we're not creating too many clients). However I would still like to refactor this to close the client once the message is published.
The linked issue recommends client.api.transport._channel.close() after you'r finished with the client. However, in this case I'm only finished with it after the pubsub_callback has been triggered.
I'm not seeing any way to get the client from the future, and the callback add_done_callback doesn't (right so) doesn't allow to send arguments.
Any creative solutions?
I need to bite the bullet anyway and refactor the heavy pubsub clients, but it's not always clear cut.
Update:
Looking at the code, it appears as if this would successfully close the client after the future:
def send_oneoff_pubsub_message(self, client=None):
if not client:
client = self.get_client('pubsubpub') # Creates a pubsub publisher client
future = client.publish({...})
try:
future.exception(timeout=10)
except Exception as exc:
print("error")
future.add_done_callback(pubsub_callback)
future.result(timeout=10)
client.api.transport._channel.close()
Any downside of such asn approach? except that the function blocks until published (which is OK for me)

Your code does a mix of asynchronous operation and synchronous operation. After you call publish, you wait up to 10 seconds for an exception on the future. Within 10 seconds in most cases, publish will have completed anyway, so at that point, you might as well call future.result synchronously and not even bother with the add_done_callback:
def pubsub_callback(future):
try:
message_id = future.result()
LOGGER.info("Successfully published %s", message_id)
except Exception as exc:
print("error")
def send_oneoff_pubsub_message(self, client=None):
if not client:
client_created = True
client = self.get_client('pubsubpub') # Creates a pubsub publisher client
future = client.publish({...})
pubsub_callback(future)
if client_created:
client.api.transport._channel.close()
If you want to do things asynchronously, you could use functools.partial:
from functools import partial
def pubsub_callback(client, future):
try:
message_id = future.result()
LOGGER.info("Successfully published %s", message_id)
except Exception as exc:
print("error")
if client:
client.api.transport._channel.close()
def send_oneoff_pubsub_message(self, client=None):
if not client:
client_created = True
client = self.get_client('pubsubpub') # Creates a pubsub publisher client
if client_created:
callback = partial(pubsub_callback, client)
else:
callback = partial(pubsub_callback, None)
future = client.publish({...})
future.add_done_callback(callback)
Either way should allow you to close the client at the desired point.

Related

Can I send an alert when a message is published to a pubsub topic?

We are using pubsub & a cloud function to process a stream of incoming data. I am setting up a dead letter topic to handle cases where a message cannot be processed, as described at Cloud Pub/Sub > Guides > Handling message failures.
I've configured a subscription on the dead-letter topic to retain messages for 7 days, we're doing this using terraform:
resource "google_pubsub_subscription" "dead_letter_monitoring" {
project = var.project_id
name = "var.dead_letter_sub_name
topic = google_pubsub_topic.dead_letter.name
expiration_policy { ttl = "" }
message_retention_duration = "604800s" # 7 days
retain_acked_messages = true
ack_deadline_seconds = 600
}
We've tested our cloud function robustly and consequently our expectation is that messages will appear on this dead-letter topic very very rarely, perhaps never. Nevertheless we're putting it in place just to make sure that we catch any anomalies.
Given the rarity of which we expect messages to appear on the dead-letter-topic we need to set up an alert to send an email when such a message appears. Is it possible to do this? I've taken a look through the alerts one can create at https://console.cloud.google.com/monitoring/alerting/policies/create however I didn't see anything that could accomplish this.
I know that I could write a cloud function to consume a message from the subscription and act upon it accordingly however I'd rather not have to do that, a monitoring alert feels like a much more elegant way of achieving this.
is this possible?
Yes, you can use Cloud Monitoring for that. Create a new policy and perform that configuration
Select PubSub Topic and Published message. Observe the value every minute and count them (aligner in the advanced option). Now, in the config, when it's above 0 from the most recent value, the alert is raised.
To filter on only your topic you can add a filter by topic_id on your topic name.
Then, configure your alert to send an email. It should work!

Creating a cluster before sending a job to dataproc programmatically

I'm trying to schedule a PySpark Job. I followed the GCP documentation and ended up deploying a little python script to App Engine which does the following :
authenticate using a service account
submit a job to a cluster
The problem is, I need the cluster to be up and running otherwise the job won't be sent (duh !) but I don't want the cluster to always be up and running, especially since my job needs to run once a month.
I wanted to add the creation of a cluster in my python script but the call is asynchronous (it makes an HTTP request) and thus my job is submitted after the cluster creation call but before the cluster is really up and running.
How could I do ?
I'd like something cleaner than just waiting for a few minutes in my script !
Thanks
EDIT : Here's what my code looks like so far :
To launch the job
class EnqueueTaskHandler(webapp2.RequestHandler):
def get(self):
task = taskqueue.add(
url='/run',
target='worker')
self.response.write(
'Task {} enqueued, ETA {}.'.format(task.name, task.eta))
app = webapp2.WSGIApplication([('/launch', EnqueueTaskHandler)], debug=True)
The job
class CronEventHandler(webapp2.RequestHandler):
def create_cluster(self, dataproc, project, zone, region, cluster_name):
zone_uri = 'https://www.googleapis.com/compute/v1/projects/{}/zones/{}'.format(project, zone)
cluster_data = {...}
dataproc.projects().regions().clusters().create(
projectId=project,
region=region,
body=cluster_data).execute()
def wait_for_cluster(self, dataproc, project, region, clustername):
print('Waiting for cluster to run...')
while True:
result = dataproc.projects().regions().clusters().get(
projectId=project,
region=region,
clusterName=clustername).execute()
# Handle exceptions
if result['status']['state'] != 'RUNNING':
time.sleep(60)
else:
return result
def wait_for_job(self, dataproc, project, region, job_id):
print('Waiting for job to finish...')
while True:
result = dataproc.projects().regions().jobs().get(
projectId=project,
region=region,
jobId=job_id).execute()
# Handle exceptions
print(result['status']['state'])
if result['status']['state'] == 'ERROR' or result['status']['state'] == 'DONE':
return result
else:
time.sleep(60)
def submit_job(self, dataproc, project, region, clusterName):
job = {...}
result = dataproc.projects().regions().jobs().submit(projectId=project,region=region,body=job).execute()
return result['reference']['jobId']
def post(self):
dataproc = googleapiclient.discovery.build('dataproc', 'v1')
project = '...'
region = "..."
zone = "..."
clusterName = '...'
self.create_cluster(dataproc, project, zone, region, clusterName)
self.wait_for_cluster(dataproc, project, region, clusterName)
job_id = self.submit_job(dataproc,project,region,clusterName)
self.wait_for_job(dataproc,project,region,job_id)
dataproc.projects().regions().clusters().delete(projectId=project, region=region, clusterName=clusterName).execute()
self.response.write("JOB SENT")
app = webapp2.WSGIApplication([('/run', CronEventHandler)], debug=True)
Everything works until the deletion of the cluster. At this point I get a "DeadlineExceededError: The overall deadline for responding to the HTTP request was exceeded." Any idea ?
In addition to general polling either through list or get requests on the Cluster or the Operation returned with the CreateCluster request, for single-use clusters like this you can also consider using the Dataproc Workflows API and possibly its InstantiateInline interface if you don't want to use full-fledged workflow templates; in this API you use a single request to specify cluster settings along with jobs to submit, and the jobs will automatically run as soon as the cluster is ready to take it, after which the cluster will be deleted automatically.
You can use the Google Cloud Dataproc API to create, delete and list clusters.
The list operation can be (repeatedly) performed after create and delete operations to confirm that they completed successfully, since it provides the ClusterStatus of the clusters in the results with the relevant State information:
UNKNOWN The cluster state is unknown.
CREATING The cluster is being created and set up. It is not ready for use.
RUNNING The cluster is currently running and healthy. It is ready for use.
ERROR The cluster encountered an error. It is not ready for use.
DELETING The cluster is being deleted. It cannot be used.
UPDATING The cluster is being updated. It continues to accept and process jobs.
To prevent plain waiting between the (repeated) list invocations (in general not a good thing to do on GAE) you can enqueue delayed tasks in a push task queue (with the relevant context information) allowing you to perform such list operations at a later time. For example, in python, see taskqueue.add():
countdown -- Time in seconds into the future that this task should run or be leased. Defaults to zero. Do not specify this argument if
you specified an eta.
eta -- A datetime.datetime that specifies the absolute earliest time at which the task should run. You cannot specify this argument if
the countdown argument is specified. This argument can be time
zone-aware or time zone-naive, or set to a time in the past. If the
argument is set to None, the default value is now. For pull tasks, no
worker can lease the task before the time indicated by the eta
argument.
If at the task execution time the result indicates the operation of interest is still in progress simply enqueue another such delayed task - effectively polling but without an actual wait/sleep.

Pidgin: cannot send messages or set topics in chats via dbus

I would like to send messages to Pidgin chats or set chat topics via dbus. Following this guide I was able to write some pretty straightforward code to do just that, and it does indeed result in messages appearing or chat topics being changed... but it only seems to affect my window, without the other participants being aware of any messages or topic changes.
I'm using
purple.PurpleConvChatSetTopic(chat_data, user, topic)
and
purple.PurpleConvChatWrite(chat_data, user, message, flag, time)
I don't think this is due to any misuse of the dbus api as the calls actually result in actions. I just wonder if I need to perform some sort of authentication first? Or maybe the user can only be the current user? I tried with my nick and also setting it as unicode but to no avail.
Here is the complete code anyway:
import dbus
import time
# define chat_name, user, topic, message
bus = dbus.SessionBus()
obj = bus.get_object('im.pidgin.purple.PurpleService', '/im/pidgin/purple/PurpleObject')
purple = dbus.Interface(obj, 'im.pidgin.purple.PurpleInterface')
for p in purple.PurpleGetConversations():
if purple.PurpleConversationGetName(p) == chat_name:
chat = p
chat_data = purple.PurpleConversationGetChatData(chat)
purple.PurpleConvChatSetTopic(chat_data, user, topic)
purple.PurpleConvChatWrite(chat_data, user, message, 0, int(time.time()))

'IntegrityError: column username is not unique' while using Django User model in test

While running some tests, I started to get an IntegrityError in my setUp function. Here is my code:
def setUp(self):
self.client = Client()
self.emplUser = User.objects.create_user('employee#email.com', 'employee#email.com', 'nothing')
self.servUser1 = User.objects.create_user('thebestcompany#email.com', 'thebestcompany#email.com', 'nothing')
self.servUser2 = User.objects.create_user('theothercompany#email.com', 'theothercompany#email.com', 'nothing')
self.custUser1 = User.objects.create_user('john#email.com', 'john#email.com', 'nothing')
self.custUser2 = User.objects.create_user('marcus#email.com', 'marcus#email.com', 'nothing')
... save users here ...
Im wondering as to how this IntegrityError keeps getting raised. I delete all the users in the tearDown function and am using sqlite3 as my DB backend. I see no conflicting usernames and in production, I have no issues with using emails as usernames.
This started happening only half an hour ago, out of the blue. Has anyone run into a solution to this problem?
I'm sure you're not suffering this problem anymore since you wrote 18 months ago, but I had this problem too, and finally figured out what was happening. When using Postgres for test cases, DB changes are done in a transaction and simply rolled back, and so it is not necessary to explicitly clear tables in tearDown(), however, in SQLite, it is necessary.
Late but more appropriate answer, for the people who would land there after a google search:
When there is interaction with the database in your tests (typically, creating model instances), you should subclass your test class from django.test.TestCase, which flushes the database after each test is run.
Then you don't need to write a tedious tearDown method in all your test classes.
See https://docs.djangoproject.com/en/dev/topics/testing/overview/#writing-tests

How to delete all datastore in Google App Engine?

Does anyone know how to delete all datastore in Google App Engine?
If you're talking about the live datastore, open the dashboard for your app (login on appengine) then datastore --> dataviewer, select all the rows for the table you want to delete and hit the delete button (you'll have to do this for all your tables).
You can do the same programmatically through the remote_api (but I never used it).
If you're talking about the development datastore, you'll just have to delete the following file: "./WEB-INF/appengine-generated/local_db.bin". The file will be generated for you again next time you run the development server and you'll have a clear db.
Make sure to clean your project afterwards.
This is one of the little gotchas that come in handy when you start playing with the Google Application Engine. You'll find yourself persisting objects into the datastore then changing the JDO object model for your persistable entities ending up with obsolete data that'll make your app crash all over the place.
The best approach is the remote API method as suggested by Nick, he's an App Engine engineer from Google, so trust him.
It's not that difficult to do, and the latest 1.2.5 SDK provides the remote_shell_api.py out of the shelf. So go to download the new SDK. Then follow the steps:
connect remote server in your commandline: remote_shell_api.py yourapp /remote_api
The shell will ask for your login info, and if authorized, will make a Python shell for you. You need setup url handler for /remote_api in your app.yaml
fetch the entities you'd like to delete, the code looks something like:
from models import Entry
query = Entry.all(keys_only=True)
entries =query.fetch(1000)
db.delete(entries)
\# This could bulk delete 1000 entities a time
Update 2013-10-28:
remote_shell_api.py has been replaced by remote_api_shell.py, and you should connect with remote_api_shell.py -s your_app_id.appspot.com, according to the documentation.
There is a new experimental feature Datastore Admin, after enabling it in app settings, you can bulk delete as well as backup your datastore through the web ui.
The fastest and efficient way to handle bulk delete on Datastore is by using the new mapper API announced on the latest Google I/O.
If your language of choice is Python, you just have to register your mapper in a mapreduce.yaml file and define a function like this:
from mapreduce import operation as op
def process(entity):
yield op.db.Delete(entity)
On Java you should have a look to this article that suggests a function like this:
#Override
public void map(Key key, Entity value, Context context) {
log.info("Adding key to deletion pool: " + key);
DatastoreMutationPool mutationPool = this.getAppEngineContext(context)
.getMutationPool();
mutationPool.delete(value.getKey());
}
EDIT:
Since SDK 1.3.8, there's a Datastore admin feature for this purpose
You can clear the development server datastore when you run the server:
/path/to/dev_appserver.py --clear_datastore=yes myapp
You can also abbreviate --clear_datastore with -c.
If you have a significant amount of data, you need to use a script to delete it. You can use remote_api to clear the datastore from the client side in a straightforward manner, though.
Here you go: Go to Datastore Admin, and then select the Entity type you want to delete and click Delete. Mapreduce will take care of deleting!
There are several ways you can use to remove entries from App Engine's Datastore:
First, think whether you really need to remove entries. This is expensive and it might be cheaper to not remove them.
You can delete all entries by hand using the Datastore Admin.
You can use the Remote API and remove entries interactively.
You can remove the entries programmatically using a couple lines of code.
You can remove them in bulk using Task Queues and Cursors.
Or you can use Mapreduce to get something more robust and fancier.
Each one of these methods is explained in the following blog post:
http://www.shiftedup.com/2015/03/28/how-to-bulk-delete-entries-in-app-engine-datastore
Hope it helps!
The zero-setup way to do this is to send an execute-arbitrary-code HTTP request to the admin service that your running app already, automatically, has:
import urllib
import urllib2
urllib2.urlopen('http://localhost:8080/_ah/admin/interactive/execute',
data = urllib.urlencode({'code' : 'from google.appengine.ext import db\n' +
'db.delete(db.Query())'}))
Source
I got this from http://code.google.com/appengine/articles/remote_api.html.
Create the Interactive Console
First, you need to define an interactive appenginge console. So, create a file called appengine_console.py and enter this:
#!/usr/bin/python
import code
import getpass
import sys
# These are for my OSX installation. Change it to match your google_appengine paths. sys.path.append("/Applications/GoogleAppEngineLauncher.app/Contents/Resources/GoogleAppEngine-default.bundle/Contents/Resources/google_appengine")
sys.path.append("/Applications/GoogleAppEngineLauncher.app/Contents/Resources/GoogleAppEngine-default.bundle/Contents/Resources/google_appengine/lib/yaml/lib")
from google.appengine.ext.remote_api import remote_api_stub
from google.appengine.ext import db
def auth_func():
return raw_input('Username:'), getpass.getpass('Password:')
if len(sys.argv) < 2:
print "Usage: %s app_id [host]" % (sys.argv[0],)
app_id = sys.argv[1]
if len(sys.argv) > 2:
host = sys.argv[2]
else:
host = '%s.appspot.com' % app_id
remote_api_stub.ConfigureRemoteDatastore(app_id, '/remote_api', auth_func, host)
code.interact('App Engine interactive console for %s' % (app_id,), None, locals())
Create the Mapper base class
Once that's in place, create this Mapper class. I just created a new file called utils.py and threw this:
class Mapper(object):
# Subclasses should replace this with a model class (eg, model.Person).
KIND = None
# Subclasses can replace this with a list of (property, value) tuples to filter by.
FILTERS = []
def map(self, entity):
"""Updates a single entity.
Implementers should return a tuple containing two iterables (to_update, to_delete).
"""
return ([], [])
def get_query(self):
"""Returns a query over the specified kind, with any appropriate filters applied."""
q = self.KIND.all()
for prop, value in self.FILTERS:
q.filter("%s =" % prop, value)
q.order("__key__")
return q
def run(self, batch_size=100):
"""Executes the map procedure over all matching entities."""
q = self.get_query()
entities = q.fetch(batch_size)
while entities:
to_put = []
to_delete = []
for entity in entities:
map_updates, map_deletes = self.map(entity)
to_put.extend(map_updates)
to_delete.extend(map_deletes)
if to_put:
db.put(to_put)
if to_delete:
db.delete(to_delete)
q = self.get_query()
q.filter("__key__ >", entities[-1].key())
entities = q.fetch(batch_size)
Mapper is supposed to be just an abstract class that allows you to iterate over every entity of a given kind, be it to extract their data, or to modify them and store the updated entities back to the datastore.
Run with it!
Now, start your appengine interactive console:
$python appengine_console.py <app_id_here>
That should start the interactive console. In it create a subclass of Model:
from utils import Mapper
# import your model class here
class MyModelDeleter(Mapper):
KIND = <model_name_here>
def map(self, entity):
return ([], [entity])
And, finally, run it (from you interactive console):
mapper = MyModelDeleter()
mapper.run()
That's it!
You can do it using the web interface. Login into your account, navigate with links on the left hand side. In Data Store management you have options to modify and delete data. Use respective options.
I've created an add-in panel that can be used with your deployed App Engine apps. It lists the kinds that are present in the datastore in a dropdown, and you can click a button to schedule "tasks" that delete all entities of a specific kind or simply everything. You can download it here:
http://code.google.com/p/jobfeed/wiki/Nuke
For Python, 1.3.8 includes an experimental admin built-in for this. They say: "enable the following builtin in your app.yaml file:"
builtins:
- datastore_admin: on
"Datastore delete is currently available only with the Python runtime. Java applications, however, can still take advantage of this feature by creating a non-default Python application version that enables Datastore Admin in the app.yaml. Native support for Java will be included in an upcoming release."
Open "Datastore Admin" for your application and enable Admin. Then all of your entities will be listed with check boxes. You can simply select the unwanted entites and delete them.
This is what you're looking for...
db.delete(Entry.all(keys_only=True))
Running a keys-only query is much faster than a full fetch, and your quota will take a smaller hit because keys-only queries are considered small ops.
Here's a link to an answer from Nick Johnson describing it further.
Below is an end-to-end REST API solution to truncating a table...
I setup a REST API to handle database transactions where routes are directly mapped through to the proper model/action. This can be called by entering the right url (example.com/inventory/truncate) and logging in.
Here's the route:
Route('/inventory/truncate', DataHandler, defaults={'_model':'Inventory', '_action':'truncate'})
Here's the handler:
class DataHandler(webapp2.RequestHandler):
#basic_auth
def delete(self, **defaults):
model = defaults.get('_model')
action = defaults.get('_action')
module = __import__('api.models', fromlist=[model])
model_instance = getattr(module, model)()
result = getattr(model_instance, action)()
It starts by loading the model dynamically (ie Inventory found under api.models), then calls the correct method (Inventory.truncate()) as specified in the action parameter.
The #basic_auth is a decorator/wrapper that provides authentication for sensitive operations (ie POST/DELETE). There's also an oAuth decorator available if you're concerned about security.
Finally, the action is called:
def truncate(self):
db.delete(Inventory.all(keys_only=True))
It looks like magic but it's actually very straightforward. The best part is, delete() can be re-used to handle deleting one-or-many results by adding another action to the model.
You can Delete All Datastore by deleting all Kinds One by One.
with google appengine dash board. Please follow these Steps.
Login to https://console.cloud.google.com/datastore/settings
Click Open Datastore Admin. (Enable it if not enabled.)
Select all Entities and press delete.(This Step run a map reduce job for deleting all selected Kinds.)
for more information see This image http://storage.googleapis.com/bnifsc/Screenshot%20from%202015-01-31%2023%3A58%3A41.png
If you have a lot of data, using the web interface could be time consuming. The App Engine Launcher utility lets you delete everything in one go with the 'Clear datastore on launch' checkbox. This utility is now available for both Windows and Mac (Python framework).
For the development server, instead of running the server through the google app engine launcher, you can run it from the terminal like:
dev_appserver.py --port=[portnumber] --clear_datastore=yes [nameofapplication]
ex: my application "reader" runs on port 15080. After modify the code and restart the server, I just run "dev_appserver.py --port=15080 --clear_datastore=yes reader".
It's good for me.
Adding answer about recent developments.
Google recently added datastore admin feature. You can backup, delete or copy your entities to another app using this console.
https://developers.google.com/appengine/docs/adminconsole/datastoreadmin#Deleting_Entities_in_Bulk
I often don't want to delete all the data store so I pull a clean copy of /war/WEB-INF/local_db.bin out source control. It may just be me but it seems even with the Dev Mode stopped I have to physically remove the file before pulling it. This is on Windows using the subversion plugin for Eclipse.
PHP variation:
import com.google.appengine.api.datastore.Query;
import com.google.appengine.api.datastore.DatastoreServiceFactory;
define('DATASTORE_SERVICE', DatastoreServiceFactory::getDatastoreService());
function get_all($kind) {
$query = new Query($kind);
$prepared = DATASTORE_SERVICE->prepare($query);
return $prepared->asIterable();
}
function delete_all($kind, $amount = 0) {
if ($entities = get_all($kind)) {
$r = $t = 0;
$delete = array();
foreach ($entities as $entity) {
if ($r < 500) {
$delete[] = $entity->getKey();
} else {
DATASTORE_SERVICE->delete($delete);
$delete = array();
$r = -1;
}
$r++; $t++;
if ($amount && $amount < $t) break;
}
if ($delete) {
DATASTORE_SERVICE->delete($delete);
}
}
}
Yes it will take time and 30 sec. is a limit. I'm thinking to put an ajax app sample to automate beyond 30 sec.
for amodel in db.Model.__subclasses__():
dela=[]
print amodel
try:
m = amodel()
mq = m.all()
print mq.count()
for mw in mq:
dela.append(mw)
db.delete(dela)
#~ print len(dela)
except:
pass
If you're using ndb, the method that worked for me for clearing the datastore:
ndb.delete_multi(ndb.Query(default_options=ndb.QueryOptions(keys_only=True)))
For any datastore that's on app engine, rather than local, you can use the new Datastore API. Here's a primer for how to get started.
I wrote a script that deletes all non-built in entities. The API is changing pretty rapidly, so for reference, I cloned it at commit 990ab5c7f2063e8147bcc56ee222836fd3d6e15b
from gcloud import datastore
from gcloud.datastore import SCOPE
from gcloud.datastore.connection import Connection
from gcloud.datastore import query
from oauth2client import client
def get_connection():
client_email = 'XXXXXXXX#developer.gserviceaccount.com'
private_key_string = open('/path/to/yourfile.p12', 'rb').read()
svc_account_credentials = client.SignedJwtAssertionCredentials(
service_account_name=client_email,
private_key=private_key_string,
scope=SCOPE)
return Connection(credentials=svc_account_credentials)
def connect_to_dataset(dataset_id):
connection = get_connection()
datastore.set_default_connection(connection)
datastore.set_default_dataset_id(dataset_id)
if __name__ == "__main__":
connect_to_dataset(DATASET_NAME)
gae_entity_query = query.Query()
gae_entity_query.keys_only()
for entity in gae_entity_query.fetch():
if entity.kind[0] != '_':
print entity.kind
entity.key.delete()
continuing the idea of svpino it is wisdom to reuse records marked as delete. (his idea was not to remove, but mark as "deleted" unused records). little bit of cache/memcache to handle working copy and write only difference of states (before and after desired task) to datastore will make it better. for big tasks it is possible to write itermediate difference chunks to datastore to avoid data loss if memcache disappeared. to make it loss-proof it is possible to check integrity/existence of memcached results and restart task (or required part) to repeat missing computations. when data difference is written to datastore, required computations are discarded in queue.
other idea similar to map reduced is to shard entity kind to several different entity kinds, so it will be collected together and visible as single entity kind to final user. entries are only marked as "deleted". when "deleted" entries amount per shard overcomes some limit, "alive" entries are distributed between other shards, and this shard is closed forever and then deleted manually from dev console (guess at less cost) upd: seems no drop table at console, only delete record-by-record at regular price.
it is possible to delete by query by chunks large set of records without gae failing (at least works locally) with possibility to continue in next attempt when time is over:
qdelete.getFetchPlan().setFetchSize(100);
while (true)
{
long result = qdelete.deletePersistentAll(candidates);
LOG.log(Level.INFO, String.format("deleted: %d", result));
if (result <= 0)
break;
}
also sometimes it useful to make additional field in primary table instead of putting candidates (related records) into separate table. and yes, field may be unindexed/serialized array with little computation cost.
For all people that need a quick solution for the dev server (as time of writing in Feb. 2016):
Stop the dev server.
Delete the target directory.
Rebuild the project.
This will wipe all data from the datastore.
I was so frustrated about existing solutions for deleting all data in the live datastore that I created a small GAE app that can delete quite some amount of data within its 30 seconds.
How to install etc: https://github.com/xamde/xydra
For java
DatastoreService db = DatastoreServiceFactory.getDatastoreService();
List<Key> keys = new ArrayList<Key>();
for(Entity e : db.prepare(new Query().setKeysOnly()).asIterable())
keys.add(e.getKey());
db.delete(keys);
Works well in Development Server
You have 2 simple ways,
#1: To save cost, delete the entire project
#2: using ts-datastore-orm:
https://www.npmjs.com/package/ts-datastore-orm
await Entity.truncate();
The truncate can delete around 1K rows per seconds
Here's how I did this naively from a vanilla Google Cloud Shell (no GAE) with python3:
from google.cloud import datastore
client = datastore.Client()
query.keys_only()
for counter, entity in enumerate(query.fetch()):
if entity.kind.startswith('_'): # skip reserved kinds
continue
print(f"{counter}: {entity.key}")
client.delete(entity.key)
This takes a very long time even with a relatively small amount of keys but it works.
More info about the Python client library: https://googleapis.dev/python/datastore/latest/client.html
As of 2022, there are two ways to delete a kind from a (largeish) datastore to the best of my knowledge. Google recommends using a Dataflow template. The template will basically pull each entity one by one subject to a GQL query, and then delete it. Interestingly, if you are deleting a large number of rows (> 10m), you will run into datastore troubles; as it will fail to provide enough capacity, and your operations to the datastore will start timing out. However, only the kind you are mass deleting from will be effected.
If you have less than 10m rows, you can just use this go script:
import (
"cloud.google.com/go/datastore"
"context"
"fmt"
"google.golang.org/api/option"
"log"
"strings"
"sync"
"time"
)
const (
batchSize = 10000 // number of keys to get in a single batch
deleteBatchSize = 500 // number of keys to delete in a single batch
projectID = "name-of-your-GCP-project"
serviceAccount = "path-to-sa-file"
table = "kind-to-delete"
)
func min(a, b int) int {
if a < b {
return a
}
return b
}
func deleteBatch(table string) int {
ctx := context.Background()
client, err := datastore.NewClient(ctx, projectID, option.WithCredentialsFile(serviceAccount))
if err != nil {
log.Fatalf("Failed to open client: %v", err)
}
defer client.Close()
query := datastore.NewQuery(table).KeysOnly().Limit(batchSize)
keys, err := client.GetAll(ctx, query, nil)
if err != nil {
fmt.Printf("%s Failed to get %d keys : %v\n", table, batchSize, err)
return -1
}
var wg sync.WaitGroup
for i := 0; i < len(keys); i += deleteBatchSize {
wg.Add(1)
go func(i int) {
batch := keys[i : i+min(len(keys)-i, deleteBatchSize)]
if err := client.DeleteMulti(ctx, batch); err != nil {
// not a big problem, we'll get them next time ;)
fmt.Printf("%s Failed to delete multi: %v", table, err)
}
wg.Done()
}(i)
}
wg.Wait()
return len(keys)
}
func main() {
var globalStartTime = time.Now()
fmt.Printf("Deleting \033[1m%s\033[0m\n", table)
for {
startTime := time.Now()
count := deleteBatch(table)
if count >= 0 {
rate := float64(count) / time.Since(startTime).Seconds()
fmt.Printf("Deleted %d keys from %s in %.2fs, rate %.2f keys/s\n", count, table, time.Since(startTime).Seconds(), rate)
if count == 0 {
fmt.Printf("%s is now clear.\n", table)
break
}
} else {
fmt.Printf("Retrying after short cooldown\n")
time.Sleep(10 * time.Second)
}
}
fmt.Printf("Total time taken %s.\n", time.Since(globalStartTime))
}

Resources