Speed up Datastore deletes by batch deletions with multithreading #2182
Conversation
Signed-off-by: Pamela Toman <ptoman@paloaltonetworks.com>
|
Hi @ptoman-pa. Thanks for your PR. I'm waiting for a feast-dev member to verify that this patch is reasonable to test. If it is, they should reply with Once the patch is verified, the new status will be reflected by the I understand the commands that are listed here. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. |
| @@ -32,6 +35,9 @@ | |||
| from feast.protos.feast.types.Value_pb2 import Value as ValueProto | |||
| from feast.repo_config import FeastConfigBaseModel, RepoConfig | |||
| from feast.usage import log_exceptions_and_usage, tracing_span | |||
| from feast.utils.generic_utils import AtomicCounter | |||
There was a problem hiding this comment.
We've generally tried to avoid adding utils modules since it's a kitchen sink/code smell. If this AtomicCounter is only used in this datastore class, maybe we can just define it here? Or rename feast.utils.generic_utils to feast.atomic or something?
There was a problem hiding this comment.
I'm actually glad to hear this :) I wanted to use the patterns of the codebase. I'm going to move it into an innerclass just for the _delete_all_values since it's only used for that logging statement (or we could remove the logging), and it can always be migrated to increasing scope if needed.
| while True: | ||
| client.delete_multi(deletion_queue.get()) | ||
| shared_counter.increment() | ||
| LOGGER.info( |
Signed-off-by: Pamela Toman <ptoman@paloaltonetworks.com>
Codecov Report
@@ Coverage Diff @@
## master #2182 +/- ##
==========================================
+ Coverage 84.59% 84.64% +0.04%
==========================================
Files 102 102
Lines 8186 8230 +44
==========================================
+ Hits 6925 6966 +41
- Misses 1261 1264 +3
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
ptoman-pa
force-pushed
the
faster-deletes
branch
from
9679440
to
67c64cf
Compare
| with self.lock: | ||
| self.value += 1 | ||
|
|
||
| BATCH_SIZE = 500 # Dec 2021: delete_multi has a max size of 500 |
There was a problem hiding this comment.
Nit is there a doc we can link to in case this changes in the future?
There was a problem hiding this comment.
I just went looking and couldn't find an obvious place. If we try to query with more than 500, the server responds:
google.api_core.exceptions.InvalidArgument: 400 cannot write more than 500 entities in a single call
There was a problem hiding this comment.
Signed-off-by: Pamela Toman <ptoman@paloaltonetworks.com>
|
/ok-to-test |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: achals, ptoman-pa The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
What this PR does / why we need it:
Speeds up teardown in Datastore by 3-4 orders of magnitude via Datastore's
delete_multiplus multithreading with a work queue. (Previous approach was one-request-per-element single-threaded deletions.)Which issue(s) this PR fixes:
No issue
Does this PR introduce a user-facing change?: