diff options
author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-21 11:54:28 +0000 |
---|---|---|
committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-21 11:54:28 +0000 |
commit | e6918187568dbd01842d8d1d2c808ce16a894239 (patch) | |
tree | 64f88b554b444a49f656b6c656111a145cbbaa28 /src/test/rgw/rgw_multi/zone_es.py | |
parent | Initial commit. (diff) | |
download | ceph-e6918187568dbd01842d8d1d2c808ce16a894239.tar.xz ceph-e6918187568dbd01842d8d1d2c808ce16a894239.zip |
Adding upstream version 18.2.2.upstream/18.2.2
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'src/test/rgw/rgw_multi/zone_es.py')
-rw-r--r-- | src/test/rgw/rgw_multi/zone_es.py | 256 |
1 files changed, 256 insertions, 0 deletions
diff --git a/src/test/rgw/rgw_multi/zone_es.py b/src/test/rgw/rgw_multi/zone_es.py new file mode 100644 index 000000000..e98b3fdd8 --- /dev/null +++ b/src/test/rgw/rgw_multi/zone_es.py @@ -0,0 +1,256 @@ +import json +import requests.compat +import logging + +import boto +import boto.s3.connection + +import dateutil.parser + +from nose.tools import eq_ as eq +from itertools import zip_longest # type: ignore + +from .multisite import * +from .tools import * + +log = logging.getLogger(__name__) + +def get_key_ver(k): + if not k.version_id: + return 'null' + return k.version_id + +def check_object_eq(k1, k2, check_extra = True): + assert k1 + assert k2 + log.debug('comparing key name=%s', k1.name) + eq(k1.name, k2.name) + eq(k1.metadata, k2.metadata) + # eq(k1.cache_control, k2.cache_control) + eq(k1.content_type, k2.content_type) + # eq(k1.content_encoding, k2.content_encoding) + # eq(k1.content_disposition, k2.content_disposition) + # eq(k1.content_language, k2.content_language) + eq(k1.etag, k2.etag) + mtime1 = dateutil.parser.parse(k1.last_modified) + mtime2 = dateutil.parser.parse(k2.last_modified) + assert abs((mtime1 - mtime2).total_seconds()) < 1 # handle different time resolution + if check_extra: + eq(k1.owner.id, k2.owner.id) + eq(k1.owner.display_name, k2.owner.display_name) + # eq(k1.storage_class, k2.storage_class) + eq(k1.size, k2.size) + eq(get_key_ver(k1), get_key_ver(k2)) + # eq(k1.encrypted, k2.encrypted) + +def make_request(conn, method, bucket, key, query_args, headers): + result = conn.make_request(method, bucket=bucket, key=key, query_args=query_args, headers=headers) + if result.status // 100 != 2: + raise boto.exception.S3ResponseError(result.status, result.reason, result.read()) + return result + + +class MDSearch: + def __init__(self, conn, bucket_name, query, query_args = None, marker = None): + self.conn = conn + self.bucket_name = bucket_name or '' + if bucket_name: + self.bucket = boto.s3.bucket.Bucket(name=bucket_name) + else: + self.bucket = None + self.query = query + self.query_args = query_args + self.max_keys = None + self.marker = marker + + def raw_search(self): + q = self.query or '' + query_args = append_query_arg(self.query_args, 'query', requests.compat.quote_plus(q)) + if self.max_keys is not None: + query_args = append_query_arg(query_args, 'max-keys', self.max_keys) + if self.marker: + query_args = append_query_arg(query_args, 'marker', self.marker) + + query_args = append_query_arg(query_args, 'format', 'json') + + headers = {} + + result = make_request(self.conn, "GET", bucket=self.bucket_name, key='', query_args=query_args, headers=headers) + + l = [] + + result_dict = json.loads(result.read()) + + for entry in result_dict['Objects']: + bucket = self.conn.get_bucket(entry['Bucket'], validate = False) + k = boto.s3.key.Key(bucket, entry['Key']) + + k.version_id = entry['Instance'] + k.etag = entry['ETag'] + k.owner = boto.s3.user.User(id=entry['Owner']['ID'], display_name=entry['Owner']['DisplayName']) + k.last_modified = entry['LastModified'] + k.size = entry['Size'] + k.content_type = entry['ContentType'] + k.versioned_epoch = entry['VersionedEpoch'] + + k.metadata = {} + for e in entry['CustomMetadata']: + k.metadata[e['Name']] = str(e['Value']) # int values will return as int, cast to string for compatibility with object meta response + + l.append(k) + + return result_dict, l + + def search(self, drain = True, sort = True, sort_key = None): + l = [] + + is_done = False + + while not is_done: + result, result_keys = self.raw_search() + + l = l + result_keys + + is_done = not (drain and (result['IsTruncated'] == "true")) + marker = result['Marker'] + + if sort: + if not sort_key: + sort_key = lambda k: (k.name, -k.versioned_epoch) + l.sort(key = sort_key) + + return l + + +class MDSearchConfig: + def __init__(self, conn, bucket_name): + self.conn = conn + self.bucket_name = bucket_name or '' + if bucket_name: + self.bucket = boto.s3.bucket.Bucket(name=bucket_name) + else: + self.bucket = None + + def send_request(self, conf, method): + query_args = 'mdsearch' + headers = None + if conf: + headers = { 'X-Amz-Meta-Search': conf } + + query_args = append_query_arg(query_args, 'format', 'json') + + return make_request(self.conn, method, bucket=self.bucket_name, key='', query_args=query_args, headers=headers) + + def get_config(self): + result = self.send_request(None, 'GET') + return json.loads(result.read()) + + def set_config(self, conf): + self.send_request(conf, 'POST') + + def del_config(self): + self.send_request(None, 'DELETE') + + +class ESZoneBucket: + def __init__(self, zone_conn, name, conn): + self.zone_conn = zone_conn + self.name = name + self.conn = conn + + self.bucket = boto.s3.bucket.Bucket(name=name) + + def get_all_versions(self): + + marker = None + is_done = False + + req = MDSearch(self.conn, self.name, 'bucket == ' + self.name, marker=marker) + + for k in req.search(): + yield k + + + + +class ESZone(Zone): + def __init__(self, name, es_endpoint, zonegroup = None, cluster = None, data = None, zone_id = None, gateways = None): + self.es_endpoint = es_endpoint + super(ESZone, self).__init__(name, zonegroup, cluster, data, zone_id, gateways) + + def is_read_only(self): + return True + + def tier_type(self): + return "elasticsearch" + + def create(self, cluster, args = None, check_retcode = True): + """ create the object with the given arguments """ + + if args is None: + args = '' + + tier_config = ','.join([ 'endpoint=' + self.es_endpoint, 'explicit_custom_meta=false' ]) + + args += [ '--tier-type', self.tier_type(), '--tier-config', tier_config ] + + return self.json_command(cluster, 'create', args, check_retcode=check_retcode) + + def has_buckets(self): + return False + + def has_roles(self): + return False + + class Conn(ZoneConn): + def __init__(self, zone, credentials): + super(ESZone.Conn, self).__init__(zone, credentials) + + def get_bucket(self, bucket_name): + return ESZoneBucket(self, bucket_name, self.conn) + + def create_bucket(self, name): + # should not be here, a bug in the test suite + log.critical('Conn.create_bucket() should not be called in ES zone') + assert False + + def check_bucket_eq(self, zone_conn, bucket_name): + assert(zone_conn.zone.tier_type() == "rados") + + log.info('comparing bucket=%s zones={%s, %s}', bucket_name, self.name, self.name) + b1 = self.get_bucket(bucket_name) + b2 = zone_conn.get_bucket(bucket_name) + + log.debug('bucket1 objects:') + for o in b1.get_all_versions(): + log.debug('o=%s', o.name) + log.debug('bucket2 objects:') + for o in b2.get_all_versions(): + log.debug('o=%s', o.name) + + for k1, k2 in zip_longest(b1.get_all_versions(), b2.get_all_versions()): + if k1 is None: + log.critical('key=%s is missing from zone=%s', k2.name, self.name) + assert False + if k2 is None: + log.critical('key=%s is missing from zone=%s', k1.name, zone_conn.name) + assert False + + check_object_eq(k1, k2) + + + log.info('success, bucket identical: bucket=%s zones={%s, %s}', bucket_name, self.name, zone_conn.name) + + return True + + def create_role(self, path, rolename, policy_document, tag_list): + assert False + + def get_conn(self, credentials): + return self.Conn(self, credentials) + + +class ESZoneConfig: + def __init__(self, cfg, section): + self.endpoint = cfg.get(section, 'endpoint') + |