TGooch44 commented on a change in pull request #277: Moving/Renaming hadoop module to filesystem URL: https://github.com/apache/incubator-iceberg/pull/277#discussion_r304655772
########## File path: python/iceberg/core/filesystem/s3_filesystem_wrapper.py ########## @@ -0,0 +1,283 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +import datetime +import io +import logging +import time +from urllib.parse import urlparse + +import boto3 +import requests +from retrying import retry +from s3fs import S3FileSystem + +from .file_status import FileStatus +from .file_system import FileSystem + +_logger = logging.getLogger(__name__) + + +S3_CLIENT = dict() +BOTO_STS_CLIENT = boto3.client('sts') + +STS_CACHED_RESPONSE = {} +REFRESH_THRESHOLD = 300 + + +@retry(wait_incrementing_start=100, wait_exponential_multiplier=4, + wait_exponential_max=5000, stop_max_delay=600000, stop_max_attempt_number=3) +def get_s3(role_arn="default", obj="resource"): + if role_arn not in S3_CLIENT: + S3_CLIENT[role_arn] = dict() + + if role_arn == "default": + S3_CLIENT["default"]["resource"] = boto3.resource('s3') + S3_CLIENT["default"]["client"] = boto3.client('s3') + else: + key, secret, token = get_sts_session_keys(role_arn) + sess = boto3.Session(aws_access_key_id=key, + aws_secret_access_key=secret, + aws_session_token=token) + S3_CLIENT[role_arn]["resource"] = sess.resource("s3") + S3_CLIENT[role_arn]["client"] = sess.client("s3") + + return S3_CLIENT.get(role_arn, dict()).get(obj) + + +def get_sts_session_keys(role_arn): + if role_arn is None: + return None, None, None + + if STS_CACHED_RESPONSE.get(role_arn) is None \ + or should_refresh_creds(role_arn): + STS_CACHED_RESPONSE[role_arn] = BOTO_STS_CLIENT.assume_role( + RoleArn=role_arn, + RoleSessionName='iceberg_python_client_{}'.format(int(time.time()))) + + return (STS_CACHED_RESPONSE[role_arn]["Credentials"]["AccessKeyId"], + STS_CACHED_RESPONSE[role_arn]["Credentials"]["SecretAccessKey"], + STS_CACHED_RESPONSE[role_arn]["Credentials"]["SessionToken"]) + + +def should_refresh_creds(role_arn): + if role_arn is None: + return False + + if STS_CACHED_RESPONSE.get(role_arn) is None: + return True + + expiration = STS_CACHED_RESPONSE[role_arn].get("Credentials", dict()).get("Expiration") + curr_dt = datetime.datetime.now(tz=datetime.timezone.utc) + + return (expiration - curr_dt).total_seconds() < REFRESH_THRESHOLD + + +class S3FileSystemWrapper(FileSystem): + _FS = dict() + + def __init__(self, conf=None, anon=False): + self.anon = anon + self.conf = conf + self.iam_role_arn = "default" + + if conf is not None: + self.iam_role_arn = conf.get("hive.aws_iam_role", "default") + + def open(self, path, mode='rb'): + if mode == "rb": + return S3File(path, self.iam_role_arn) + + elif mode == "w": + s3fs_obj = S3FileSystem() + return s3fs_obj.open(S3FileSystemWrapper.fix_s3_path(path), mode=mode) + + def delete(self, path): + raise NotImplementedError() + + def stat(self, path): + is_dir = False + st = dict() + try: + st = S3FileSystemWrapper.info(S3FileSystemWrapper.fix_s3_path(path), refresh=True) + except RuntimeError: + # must be a directory or subsequent du will fail + du = S3FileSystemWrapper.du(S3FileSystemWrapper.fix_s3_path(path), refresh=True) + if len(du) > 0: + is_dir = True + length = 0 + for key, val in du.items(): + length += val + + return FileStatus(path=path, length=st.get("Size", length), is_dir=is_dir, + blocksize=None, modification_time=st.get("LastModified"), access_time=None, + permission=None, owner=None, group=None) + + @staticmethod + def full_file(url, role_arn="default", use_requests=True): + if use_requests: + url = S3FileSystemWrapper.get_presigned_urls([url])[0] + stream = requests.get(url, stream=True).content + else: + stream = S3File(url, role_arn).read() + + return io.BytesIO(stream) + + @staticmethod + def get_presigned_urls(urls, role_arn="default"): + signed_urls = [] + for url in urls: + parsed_url = urlparse(S3FileSystemWrapper.fix_s3_path(url)) + signed_urls.append(get_s3(role_arn, "client").generate_presigned_url(ClientMethod='get_object', + Params={'Bucket': parsed_url.netloc, + 'Key': parsed_url.path[1:]})) + return signed_urls + + @staticmethod + def du(url): + raise NotImplementedError() + + @staticmethod + def info(url): + raise NotImplementedError() + + @staticmethod + def fix_s3_path(path): + if path.startswith("s3n"): + path = path.replace("s3n://", "s3://") + return path + + +class S3File(object): + MAX_CHUNK_SIZE = 4 * 1048576 + MIN_CHUNK_SIZE = 2 * 65536 Review comment: Parquet reader grabs 64k chunks of data. Performance was better for larger data pulls, so that more of these hit the already read-in buffer. 4mb seemed to get the most benefit without reading too much. On some small columns though, the whole column chunk is pretty small for the row-group and it did better when this was set to something closer to the size of the column chunk. Right now, the chunk_size it's getting set in the parquet reader if the file object that's passed in has the buffer_remote_reads attribute. There's probably some better ways of going about this, I'd love to hear your thoughts. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
