datatracker/ietf/utils/storage.py
Robert Sparks 997239a2ea
feat: write objects to blob storage (#8557)
* feat: basic blobstore infrastructure for dev

* refactor: (broken) attempt to put minio console behind nginx

* feat: initialize blobstore with boto3

* fix: abandon attempt to proxy minio. Use docker compose instead.

* feat: beginning of blob writes

* feat: storage utilities

* feat: test buckets

* chore: black

* chore: remove unused import

* chore: avoid f string when not needed

* fix: inform all settings files about blobstores

* fix: declare types for some settings

* ci: point to new target base

* ci: adjust test workflow

* fix: give the tests debug environment a blobstore

* fix: "better" name declarations

* ci: use devblobstore container

* chore: identify places to write to blobstorage

* chore: remove unreachable code

* feat: store materials

* feat: store statements

* feat: store status changes

* feat: store liaison attachments

* feat: store agendas provided with Interim session requests

* chore: capture TODOs

* feat: store polls and chatlogs

* chore: remove unneeded TODO

* feat: store drafts on submit and post

* fix: handle storage during doc expiration and resurrection

* fix: mirror an unlink

* chore: add/refine TODOs

* feat: store slide submissions

* fix: structure slide test correctly

* fix: correct sense of existence check

* feat: store some indexes

* feat: BlobShadowFileSystemStorage

* feat: shadow floorplans / host logos to the blob

* chore: remove unused import

* feat: strip path from blob shadow names

* feat: shadow photos / thumbs

* refactor: combine photo and photothumb blob kinds

The photos / thumbs were already dropped in the same
directory, so let's not add a distinction at this point.

* style: whitespace

* refactor: use kwargs consistently

* chore: migrations

* refactor: better deconstruct(); rebuild migrations

* fix: use new class in mack patch

* chore: add TODO

* feat: store group index documents

* chore: identify more TODO

* feat: store reviews

* fix: repair merge

* chore: remove unnecessary TODO

* feat: StoredObject metadata

* fix: deburr some debugging code

* fix: only set the deleted timestamp once

* chore: correct typo

* fix: get_or_create vs get and test

* fix: avoid the questionable is_seekable helper

* chore: capture future design consideration

* chore: blob store cfg for k8s

* chore: black

* chore: copyright

* ci: bucket name prefix option + run Black

Adds/uses DATATRACKER_BLOB_STORE_BUCKET_PREFIX option. Other changes
are just Black styling.

* ci: fix typo in bucket name expression

* chore: parameters in app-configure-blobstore

Allows use with other blob stores.

* ci: remove verify=False option

* fix: don't return value from __init__

* feat: option to log timing of S3Storage calls

* chore: units

* fix: deleted->null when storing a file

* style: Black

* feat: log as JSON; refactor to share code; handle exceptions

* ci: add ietf_log_blob_timing option for k8s

* test: --no-manage-blobstore option for running tests

* test: use blob store settings from env, if set

* test: actually set a couple more storage opts

* feat: offswitch (#8541)

* feat: offswitch

* fix: apply ENABLE_BLOBSTORAGE to BlobShadowFileSystemStorage behavior

* chore: log timing of blob reads

* chore: import Config from botocore.config

* chore(deps): import boto3-stubs / botocore

botocore is implicitly imported, but make it explicit
since we refer to it directly

* chore: drop type annotation that mypy loudly ignores

* refactor: add storage methods via mixin

Shares code between Document and DocHistory without
putting it in the base DocumentInfo class, which
lacks the name field. Also makes mypy happy.

* feat: add timeout / retry limit to boto client

* ci: let k8s config the timeouts via env

* chore: repair merge resolution typo

* chore: tweak settings imports

* chore: simplify k8s/settings_local.py imports

---------

Co-authored-by: Jennifer Richards <jennifer@staff.ietf.org>
2025-02-19 17:41:10 -06:00

57 lines
1.9 KiB
Python

# Copyright The IETF Trust 2020-2025, All Rights Reserved
"""Django Storage classes"""
from pathlib import Path
from django.conf import settings
from django.core.files.storage import FileSystemStorage
from ietf.doc.storage_utils import store_file
from .log import log
class NoLocationMigrationFileSystemStorage(FileSystemStorage):
def deconstruct(self):
path, args, kwargs = super().deconstruct()
kwargs["location"] = None # don't record location in migrations
return path, args, kwargs
class BlobShadowFileSystemStorage(NoLocationMigrationFileSystemStorage):
"""FileSystemStorage that shadows writes to the blob store as well
Strips directories from the filename when naming the blob.
"""
def __init__(
self,
*, # disallow positional arguments
kind: str,
location=None,
base_url=None,
file_permissions_mode=None,
directory_permissions_mode=None,
):
self.kind = kind
super().__init__(
location, base_url, file_permissions_mode, directory_permissions_mode
)
def save(self, name, content, max_length=None):
# Write content to the filesystem - this deals with chunks, etc...
saved_name = super().save(name, content, max_length)
if settings.ENABLE_BLOBSTORAGE:
# Retrieve the content and write to the blob store
blob_name = Path(saved_name).name # strips path
try:
with self.open(saved_name, "rb") as f:
store_file(self.kind, blob_name, f, allow_overwrite=True)
except Exception as err:
log(f"Failed to shadow {saved_name} at {self.kind}:{blob_name}: {err}")
return saved_name # includes the path!
def deconstruct(self):
path, args, kwargs = super().deconstruct()
kwargs["kind"] = "" # don't record "kind" in migrations
return path, args, kwargs