def download_one_backup(self, *, transfer, basebackup_data_file,
progress_callback, site):
dl_dir = os.path.join(
self.config["backup_location"],
self.config["backup_sites"][site]["prefix"],
"basebackup_incoming",
)
compat.makedirs(dl_dir, exist_ok=True)
tmp = tempfile.NamedTemporaryFile(dir=dl_dir,
prefix="basebackup.",
suffix=".pghoard")
try:
metadata = transfer.get_contents_to_fileobj(
key=basebackup_data_file,
fileobj_to_store_to=tmp,
progress_callback=progress_callback)
progress_callback(1, 1)
self.log.info("Downloaded %r", basebackup_data_file)
tmp.seek(0)
except: # pylint: disable=bare-except
self.log.exception("Problem downloading a backup file: %r",
basebackup_data_file)
tmp.close()
raise
return tmp, metadata
def setup_method(self, method):
super().setup_method(method)
self.config = self.config_template()
self.config["backup_sites"][self.test_site].update({
"basebackup_count":
1,
"basebackup_interval_hours":
1,
"nodes": [{
"host": "127.0.0.4"
}],
})
config_path = os.path.join(self.temp_dir, "pghoard.json")
write_json_file(config_path, self.config)
compat.makedirs(self.config["alert_file_dir"], exist_ok=True)
self.pghoard = PGHoard(config_path)
# This is the "final storage location" when using "local" storage type
self.local_storage_dir = os.path.join(
self.config["backup_sites"][self.test_site]["object_storage"]
["directory"], self.test_site)
self.real_check_pg_server_version = self.pghoard.check_pg_server_version
self.pghoard.check_pg_server_version = Mock(return_value=90404)
self.real_check_pg_versions_ok = self.pghoard.check_pg_versions_ok
self.pghoard.check_pg_versions_ok = Mock(return_value=True)
def _extract_pghoard_bb_v1(self, fileobj, pgdata, tablespaces):
directories = []
tar_meta = None
# | in mode to use tarfile's internal stream buffer manager, currently required because our SnappyFile
# interface doesn't do proper buffering for reads
with tarfile.open(fileobj=fileobj, mode="r|",
bufsize=IO_BLOCK_SIZE) as tar:
for tarinfo in tar:
if tarinfo.name == ".pghoard_tar_metadata.json":
tar_meta_bytes = tar.extractfile(tarinfo).read()
tar_meta = json.loads(tar_meta_bytes.decode("utf-8"))
continue
if tarinfo.name == "pgdata" or tarinfo.name == "tablespaces":
continue # ignore these directory entries
if tarinfo.name.startswith("pgdata/"):
target_name = os.path.join(pgdata, tarinfo.name[7:])
elif tarinfo.name.startswith("tablespaces/"):
tscomponents = tarinfo.name.split("/", 2)
tsname = tscomponents[1]
tspath = tablespaces[tsname]["path"]
if len(tscomponents) == 2 and tarinfo.isdir():
# Create tablespace entry
assert tar_meta["tablespaces"][tsname][
"oid"] == tablespaces[tsname]["oid"]
linkname = os.path.join(
pgdata, "pg_tblspc",
str(tablespaces[tsname]["oid"]))
os.symlink(tspath, linkname)
directories.append([tspath, tarinfo])
continue
target_name = os.path.join(tspath, tscomponents[2])
else:
raise Exception("Unrecognized path {!r} in tar".format(
tarinfo.name))
if tarinfo.isdir():
directories.append([target_name, tarinfo])
compat.makedirs(target_name, exist_ok=True)
elif tarinfo.isreg():
target_dir = os.path.dirname(target_name)
if not os.path.exists(target_dir):
compat.makedirs(target_dir, exist_ok=True)
tar.makefile(tarinfo, target_name)
tar.chmod(tarinfo, target_name)
tar.utime(tarinfo, target_name)
elif tarinfo.issym():
os.symlink(tarinfo.linkname, target_name)
else:
raise Exception(
"Unrecognized file type for file {!r} in tar".format(
tarinfo.name))
for target_name, tarinfo in directories:
tar.chmod(tarinfo, target_name)
tar.utime(tarinfo, target_name)
def _extract_pghoard_bb_v1(self, fileobj, pgdata, tablespaces):
directories = []
tar_meta = None
# | in mode to use tarfile's internal stream buffer manager, currently required because our SnappyFile
# interface doesn't do proper buffering for reads
with tarfile.open(fileobj=fileobj, mode="r|", bufsize=IO_BLOCK_SIZE) as tar:
for tarinfo in tar:
if tarinfo.name == ".pghoard_tar_metadata.json":
tar_meta_bytes = tar.extractfile(tarinfo).read()
tar_meta = json.loads(tar_meta_bytes.decode("utf-8"))
continue
if tarinfo.name == "pgdata" or tarinfo.name == "tablespaces":
continue # ignore these directory entries
if tarinfo.name.startswith("pgdata/"):
target_name = os.path.join(pgdata, tarinfo.name[7:])
elif tarinfo.name.startswith("tablespaces/"):
tscomponents = tarinfo.name.split("/", 2)
tsname = tscomponents[1]
tspath = tablespaces[tsname]["path"]
if len(tscomponents) == 2 and tarinfo.isdir():
# Create tablespace entry
assert tar_meta["tablespaces"][tsname]["oid"] == tablespaces[tsname]["oid"]
linkname = os.path.join(pgdata, "pg_tblspc", str(tablespaces[tsname]["oid"]))
os.symlink(tspath, linkname)
directories.append([tspath, tarinfo])
continue
target_name = os.path.join(tspath, tscomponents[2])
else:
raise Exception("Unrecognized path {!r} in tar".format(tarinfo.name))
if tarinfo.isdir():
directories.append([target_name, tarinfo])
compat.makedirs(target_name, exist_ok=True)
elif tarinfo.isreg():
target_dir = os.path.dirname(target_name)
if not os.path.exists(target_dir):
compat.makedirs(target_dir, exist_ok=True)
tar.makefile(tarinfo, target_name)
tar.chmod(tarinfo, target_name)
tar.utime(tarinfo, target_name)
elif tarinfo.issym():
os.symlink(tarinfo.linkname, target_name)
else:
raise Exception("Unrecognized file type for file {!r} in tar".format(tarinfo.name))
for target_name, tarinfo in directories:
tar.chmod(tarinfo, target_name)
tar.utime(tarinfo, target_name)
def config_template(self, override=None):
# NOTE: we set pg_receivexlog_path and pg_basebackup_path per site and globally mostly to verify that
# it works, the config keys are deprecated and will be removed in a future release at which point we'll
# switch to using pg_bin_directory config.
bindir, ver = find_pg_binary("")
if hasattr(psycopg2.extras, "PhysicalReplicationConnection"):
active_backup_mode = "walreceiver"
else:
active_backup_mode = "pg_receivexlog"
# Instantiate a fake PG data directory
pg_data_directory = os.path.join(str(self.temp_dir),
"PG_DATA_DIRECTORY")
os.makedirs(pg_data_directory)
open(os.path.join(pg_data_directory, "PG_VERSION"), "w").write(ver)
config = {
"alert_file_dir": os.path.join(str(self.temp_dir), "alerts"),
"backup_location": os.path.join(str(self.temp_dir), "backupspool"),
"backup_sites": {
self.test_site: {
"active_backup_mode": active_backup_mode,
"object_storage": {
"storage_type": "local",
"directory": os.path.join(self.temp_dir, "backups"),
},
"pg_data_directory": pg_data_directory,
"pg_receivexlog_path":
os.path.join(bindir, "pg_receivexlog"),
},
},
"json_state_file_path": os.path.join(self.temp_dir, "state.json"),
"pg_basebackup_path": os.path.join(bindir, "pg_basebackup"),
}
if ver == "10":
config["backup_sites"][
self.test_site]["pg_receivexlog_path"] = os.path.join(
bindir, "pg_receivewal")
if override:
all_site_overrides = override.pop("backup_sites", None)
for site_name, site_override in (all_site_overrides or {}).items():
if site_name in config["backup_sites"]:
config["backup_sites"][site_name].update(site_override)
else:
config["backup_sites"][site_name] = site_override
config.update(override)
compat.makedirs(config["alert_file_dir"], exist_ok=True)
return set_and_check_config_defaults(config)
def config_template(self, override=None):
# NOTE: we set pg_receivexlog_path and pg_basebackup_path per site and globally mostly to verify that
# it works, the config keys are deprecated and will be removed in a future release at which point we'll
# switch to using pg_bin_directory config.
bindir, ver = find_pg_binary("")
if hasattr(psycopg2.extras, "PhysicalReplicationConnection"):
active_backup_mode = "walreceiver"
else:
active_backup_mode = "pg_receivexlog"
# Instantiate a fake PG data directory
pg_data_directory = os.path.join(str(self.temp_dir), "PG_DATA_DIRECTORY")
os.makedirs(pg_data_directory)
open(os.path.join(pg_data_directory, "PG_VERSION"), "w").write(ver)
config = {
"alert_file_dir": os.path.join(str(self.temp_dir), "alerts"),
"backup_location": os.path.join(str(self.temp_dir), "backupspool"),
"backup_sites": {
self.test_site: {
"active_backup_mode": active_backup_mode,
"object_storage": {
"storage_type": "local",
"directory": os.path.join(self.temp_dir, "backups"),
},
"pg_data_directory": pg_data_directory,
"pg_receivexlog_path": os.path.join(bindir, "pg_receivexlog"),
},
},
"json_state_file_path": os.path.join(self.temp_dir, "state.json"),
"pg_basebackup_path": os.path.join(bindir, "pg_basebackup"),
}
if ver in ("10", "11"):
config["backup_sites"][self.test_site]["pg_receivexlog_path"] = os.path.join(bindir, "pg_receivewal")
if override:
all_site_overrides = override.pop("backup_sites", None)
for site_name, site_override in (all_site_overrides or {}).items():
if site_name in config["backup_sites"]:
config["backup_sites"][site_name].update(site_override)
else:
config["backup_sites"][site_name] = site_override
config.update(override)
compat.makedirs(config["alert_file_dir"], exist_ok=True)
return set_and_check_config_defaults(config)
def setup_method(self, method):
super().setup_method(method)
self.config = self.config_template()
self.config["backup_sites"][self.test_site].update({
"basebackup_count": 1,
"basebackup_interval_hours": 1,
"nodes": [{"host": "127.0.0.4"}],
})
config_path = os.path.join(self.temp_dir, "pghoard.json")
write_json_file(config_path, self.config)
compat.makedirs(self.config["alert_file_dir"], exist_ok=True)
self.pghoard = PGHoard(config_path)
# This is the "final storage location" when using "local" storage type
self.local_storage_dir = os.path.join(self.config["backup_sites"][self.test_site]["object_storage"]["directory"],
self.test_site)
self.real_check_pg_server_version = self.pghoard.check_pg_server_version
self.pghoard.check_pg_server_version = Mock(return_value=90404)
self.real_check_pg_versions_ok = self.pghoard.check_pg_versions_ok
self.pghoard.check_pg_versions_ok = Mock(return_value=True)
def _test_storage(st, driver, tmpdir, storage_config):
scratch = tmpdir.join("scratch")
compat.makedirs(str(scratch), exist_ok=True)
# File not found cases
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_metadata_for_key("NONEXISTENT")
with pytest.raises(errors.FileNotFoundFromStorageError):
st.delete_key("NONEXISTENT")
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_contents_to_file("NONEXISTENT", str(scratch.join("a")))
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_contents_to_fileobj("NONEXISTENT", BytesIO())
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_contents_to_string("NONEXISTENT")
assert st.list_path("") == []
assert st.list_path("NONEXISTENT") == []
st.store_file_from_memory("NONEXISTENT-a/x1", b"dummy", None)
dummy_file = str(scratch.join("a"))
with open(dummy_file, "wb") as fp:
fp.write(b"dummy")
st.store_file_from_disk("NONEXISTENT-b/x1", dummy_file, None)
st.store_file_from_disk("NONEXISTENT-b/x1", dummy_file, {"x": 1})
st.delete_key("NONEXISTENT-b/x1")
st.delete_key("NONEXISTENT-a/x1")
# Other basic cases
from_disk_file = str(scratch.join("a"))
input_data = b"from disk"
if driver in ["local", "sftp"]:
input_data = input_data * 150000
with open(from_disk_file, "wb") as fp:
fp.write(input_data)
st.store_file_from_disk("test1/x1", from_disk_file, None)
out = BytesIO()
reported_positions = []
def progress_callback(pos, total):
reported_positions.append((pos, total))
assert st.get_contents_to_fileobj(
"test1/x1", out, progress_callback=progress_callback) == {}
assert out.getvalue() == input_data
if driver in ["local", "sftp"]:
input_size = len(input_data)
assert reported_positions[-1] == (input_size, input_size)
if driver == "local":
assert reported_positions == [(1024 * 1024, input_size),
(input_size, input_size)]
if driver == "s3":
response = st.s3_client.head_object(
Bucket=st.bucket_name,
Key=st.format_key_for_backend("test1/x1"),
)
assert bool(response.get("ServerSideEncryption")) == bool(
storage_config.get("encrypted"))
st.store_file_from_memory("test1/x1", b"dummy", {"k": "v"})
out = BytesIO()
assert st.get_contents_to_fileobj("test1/x1", out) == {"k": "v"}
assert out.getvalue() == b"dummy"
# sftp does not support remote copy
if driver != "sftp":
# Copy file
st.copy_file(source_key="test1/x1", destination_key="test_copy/copy1")
assert st.get_contents_to_string("test_copy/copy1") == (b"dummy", {
"k": "v"
})
st.copy_file(source_key="test1/x1",
destination_key="test_copy/copy2",
metadata={"new": "meta"})
assert st.get_contents_to_string("test_copy/copy2") == (b"dummy", {
"new": "meta"
})
st.store_file_from_memory("test1/x1", b"l", {"fancymetadata": "value"})
assert st.get_contents_to_string("test1/x1") == (b"l", {
"fancymetadata": "value"
})
st.store_file_from_memory("test1/x1", b"1", None)
assert st.get_contents_to_string("test1/x1") == (b"1", {})
st.store_file_from_memory("test1/td", b"to disk", {"to-disk": "42"})
to_disk_file = str(scratch.join("b"))
assert st.get_contents_to_file("test1/td", to_disk_file) == {
"to-disk": "42"
}
created_keys = {"test1/x1", "test1/td"}
if driver == "s3":
response = st.s3_client.head_object(
Bucket=st.bucket_name,
Key=st.format_key_for_backend("test1/x1"),
)
assert bool(response.get("ServerSideEncryption")) == bool(
storage_config.get("encrypted"))
assert st.list_path("") == [
] # nothing at top level (directories not listed)
if driver == "local":
# create a dot-file (hidden), this must be ignored
target_file = os.path.join(st.prefix, "test1/.null")
with open(target_file, "w"):
pass
tlist = st.list_path("test1")
assert len(tlist) == 2
for fe in tlist:
assert isinstance(fe["last_modified"], datetime.datetime)
assert fe["last_modified"].tzinfo is not None
if fe["name"] == "test1/x1":
assert fe["size"] == 1
assert fe["metadata"] == {}
elif fe["name"] == "test1/td":
assert fe["size"] == len(b"to disk")
assert fe["metadata"] == {"to-disk": "42"}
else:
assert 0, "unexpected name in directory"
assert set(st.iter_prefixes("test1")) == set()
for key in [
"test1/sub1/sub1.1", "test1/sub2/sub2.1/sub2.1.1", "test1/sub3"
]:
st.store_file_from_memory(key, b"1", None)
created_keys.add(key)
if driver == "local":
# sub3 is a file. Actual object storage systems support this, but a file system does not
with pytest.raises(NotADirectoryError):
st.store_file_from_memory("test1/sub3/sub3.1/sub3.1.1", b"1", None)
elif driver == "sftp":
# sub3 is a file. Actual object storage systems support this, but a file system does not
with pytest.raises(errors.StorageError):
st.store_file_from_memory("test1/sub3/sub3.1/sub3.1.1", b"1", None)
else:
st.store_file_from_memory("test1/sub3/sub3.1/sub3.1.1", b"1", None)
created_keys.add("test1/sub3/sub3.1/sub3.1.1")
if driver in ["local", "sftp"]:
assert set(st.iter_prefixes("test1")) == {"test1/sub1", "test1/sub2"}
else:
assert set(st.iter_prefixes("test1")) == {
"test1/sub1", "test1/sub2", "test1/sub3"
}
assert {item["name"]
for item in st.list_path("test1")
} == {"test1/x1", "test1/td", "test1/sub3"}
assert set(st.iter_prefixes("test1/sub1")) == set()
assert {item["name"]
for item in st.list_path("test1/sub1")} == {"test1/sub1/sub1.1"}
assert {item["name"] for item in st.list_path("test1/sub2")} == set()
assert {item["name"] for item in st.list_path("test1/sub3")} == set()
assert set(st.iter_prefixes("test1/sub2")) == {"test1/sub2/sub2.1"}
if driver in ["local", "sftp"]:
assert set(st.iter_prefixes("test1/sub3")) == set() # sub3 is a file
else:
assert set(st.iter_prefixes("test1/sub3")) == {"test1/sub3/sub3.1"}
assert set(st.iter_prefixes("test1/sub3/3.1")) == set()
expected_deep_iter_test1_names = {
"test1/x1",
"test1/td",
"test1/sub1/sub1.1",
"test1/sub2/sub2.1/sub2.1.1",
"test1/sub3",
}
if driver not in ["local", "sftp"]:
expected_deep_iter_test1_names.add("test1/sub3/sub3.1/sub3.1.1")
assert {item["name"]
for item in st.list_path("test1", deep=True)
} == expected_deep_iter_test1_names
def _object_names(iterable):
names = set()
for item in iterable:
assert item.type == KEY_TYPE_OBJECT
names.add(item.value["name"])
return names
deep_names_with_key = _object_names(
st.iter_key("test1/sub3", deep=True, include_key=True))
deep_names_without_key = _object_names(
st.iter_key("test1/sub3", deep=True, include_key=False))
if driver in ["local", "sftp"]:
assert deep_names_with_key == {"test1/sub3"}
assert deep_names_without_key == set()
else:
assert deep_names_with_key == {
"test1/sub3", "test1/sub3/sub3.1/sub3.1.1"
}
assert deep_names_without_key == {"test1/sub3/sub3.1/sub3.1.1"}
if driver == "google":
# test extra props for cacheControl in google
st.store_file_from_memory("test1/x1",
b"no cache test",
metadata={"test": "value"},
extra_props={"cacheControl": "no-cache"})
if driver == "local":
# test LocalFileIsRemoteFileError for local storage
target_file = os.path.join(st.prefix, "test1/x1")
with pytest.raises(errors.LocalFileIsRemoteFileError):
st.store_file_from_disk("test1/x1", target_file, {"local": True})
assert st.get_contents_to_string("test1/x1") == (b"1", {
"local": "True"
})
with pytest.raises(errors.LocalFileIsRemoteFileError):
st.get_contents_to_file("test1/x1", target_file)
# Missing metadata is an error situation that should fail
os.unlink(target_file + ".metadata")
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_metadata_for_key("test1/x1")
for key in created_keys:
st.delete_key(key)
assert st.list_path("test1") == [] # empty again
for name in [
"test2/foo", "test2/suba/foo", "test2/subb/bar",
"test2/subb/subsub/zob"
]:
st.store_file_from_memory(name, b"somedata")
names = sorted(item["name"] for item in st.list_path("test2", deep=True))
assert names == [
"test2/foo", "test2/suba/foo", "test2/subb/bar",
"test2/subb/subsub/zob"
]
st.delete_tree("test2")
assert st.list_path("test2", deep=True) == []
test_hash = hashlib.sha256()
test_file = str(scratch.join("30m"))
test_size_send = 0
with open(test_file, "wb") as fp:
chunk = b"30m file" * 10000
while test_size_send < 30 * 1024 * 1024:
test_hash.update(chunk)
fp.write(chunk)
test_size_send += len(chunk)
test_hash_send = test_hash.hexdigest()
st.store_file_from_disk("test1/30m",
test_file,
multipart=True,
metadata={
"thirtymeg": "data",
"size": test_size_send,
"key": "value-with-a-hyphen"
})
os.unlink(test_file)
expected_meta = {
"thirtymeg": "data",
"size": str(test_size_send),
"key": "value-with-a-hyphen"
}
meta = st.get_metadata_for_key("test1/30m")
assert meta == expected_meta
progress_reports = []
def dl_progress(current_pos, expected_max):
progress_reports.append((current_pos, expected_max))
with open(test_file, "wb") as fp:
assert st.get_contents_to_fileobj(
"test1/30m", fp, progress_callback=dl_progress) == expected_meta
assert len(progress_reports) > 0
assert progress_reports[-1][0] == progress_reports[-1][1]
test_hash = hashlib.sha256()
test_size_rec = 0
with open(test_file, "rb") as fp:
while True:
chunk = fp.read(1024 * 1024)
if not chunk:
break
test_hash.update(chunk)
test_size_rec += len(chunk)
test_hash_rec = test_hash.hexdigest()
assert test_hash_rec == test_hash_send
assert test_size_rec == test_size_send
tlist = st.list_path("test1")
assert len(tlist) == 1
assert tlist[0]["name"] == "test1/30m"
assert tlist[0]["size"] == test_size_rec
if driver == "swift":
segments = test_size_send // st.segment_size
segment_list = st.list_path("test1_segments/30m")
assert len(segment_list) >= segments
if segments >= 2:
# reupload a file with the same name but with less chunks
os.truncate(test_file, st.segment_size + 1)
test_size_send = os.path.getsize(test_file)
st.store_file_from_disk("test1/30m",
test_file,
multipart=True,
metadata={
"30m": "less data",
"size": test_size_send
})
segment_list = st.list_path("test1_segments/30m")
assert len(segment_list) == 2
assert len(st.list_path("test1")) == 1
st.delete_key("test1/30m")
assert st.list_path("test1") == []
if driver == "swift":
assert st.list_path("test1_segments/30m") == []
progress_reports = []
def upload_progress(progress):
progress_reports.append(progress)
for seekable in (False, True):
for size in (300, 3 * 1024 * 1024, 11 * 1024 * 1024):
progress_reports = []
rds = RandomDataSource(size)
if seekable:
fd = BytesIO(rds.data)
else:
fd = rds
key = "test1/{}b".format(size)
st.store_file_object(key, fd, upload_progress_fn=upload_progress)
# Progress may be reported after each chunk and chunk size depends on available memory
# on current machine so there's no straightforward way of checking reasonable progress
# updates were made. Just ensure they're ordered correctly if something was provided
assert sorted(progress_reports) == progress_reports
bio = BytesIO()
st.get_contents_to_fileobj(key, bio)
buffer = bio.getbuffer()
assert len(buffer) == size
assert buffer == rds.data
st.delete_key(key)
def test_basebackups_tablespaces(self, capsys, db, pghoard, tmpdir):
# Create a test tablespace for this instance, but make sure we drop it at the end of the test as the
# database we use is shared by all test cases, and tablespaces are a global concept so the test
# tablespace could interfere with other tests
tspath = tmpdir.join("extra-ts").strpath
os.makedirs(tspath)
conn_str = pgutil.create_connection_string(db.user)
conn = psycopg2.connect(conn_str)
conn.autocommit = True
cursor = conn.cursor()
cursor.execute("CREATE TABLESPACE tstest LOCATION %s", [tspath])
r_db, r_conn = None, None
try:
cursor.execute(
"CREATE TABLE tstest (id BIGSERIAL PRIMARY KEY, value BIGINT) TABLESPACE tstest"
)
cursor.execute(
"INSERT INTO tstest (value) SELECT * FROM generate_series(1, 1000)"
)
cursor.execute("CHECKPOINT")
cursor.execute(
"SELECT oid, pg_tablespace_location(oid) FROM pg_tablespace WHERE spcname = 'tstest'"
)
res = cursor.fetchone()
assert res[1] == tspath
# Start receivexlog since we want the WALs to be able to restore later on
wal_directory = os.path.join(pghoard.config["backup_location"],
pghoard.test_site, "xlog_incoming")
makedirs(wal_directory, exist_ok=True)
pghoard.receivexlog_listener(pghoard.test_site, db.user,
wal_directory)
if conn.server_version >= 100000:
cursor.execute("SELECT txid_current(), pg_switch_wal()")
else:
cursor.execute("SELECT txid_current(), pg_switch_xlog()")
self._test_create_basebackup(capsys, db, pghoard, "local-tar")
if conn.server_version >= 100000:
cursor.execute("SELECT txid_current(), pg_switch_wal()")
cursor.execute("SELECT txid_current(), pg_switch_wal()")
else:
cursor.execute("SELECT txid_current(), pg_switch_xlog()")
cursor.execute("SELECT txid_current(), pg_switch_xlog()")
backup_out = tmpdir.join("test-restore").strpath
backup_ts_out = tmpdir.join("test-restore-tstest").strpath
# Tablespaces are extracted to their previous absolute paths by default, but the path must be empty
# and it isn't as it's still used by the running PG
with pytest.raises(RestoreError) as excinfo:
Restore().run([
"get-basebackup",
"--config",
pghoard.config_path,
"--site",
pghoard.test_site,
"--target-dir",
backup_out,
])
assert "Tablespace 'tstest' target directory" in str(excinfo.value)
assert "not empty" in str(excinfo.value)
# We can't restore tablespaces to non-existent directories either
with pytest.raises(RestoreError) as excinfo:
Restore().run([
"get-basebackup",
"--config",
pghoard.config_path,
"--site",
pghoard.test_site,
"--target-dir",
backup_out,
"--tablespace-dir",
"tstest={}".format(backup_ts_out),
])
assert "Tablespace 'tstest' target directory" in str(excinfo.value)
assert "does not exist" in str(excinfo.value)
os.makedirs(backup_ts_out)
# We can't restore if the directory isn't writable
os.chmod(backup_ts_out, 0o500)
with pytest.raises(RestoreError) as excinfo:
Restore().run([
"get-basebackup",
"--config",
pghoard.config_path,
"--site",
pghoard.test_site,
"--target-dir",
backup_out,
"--tablespace-dir",
"tstest={}".format(backup_ts_out),
])
assert "Tablespace 'tstest' target directory" in str(excinfo.value)
assert "empty, but not writable" in str(excinfo.value)
os.chmod(backup_ts_out, 0o700)
# We can't proceed if we request mappings for non-existent tablespaces
backup_other_out = tmpdir.join("test-restore-other").strpath
os.makedirs(backup_other_out)
with pytest.raises(RestoreError) as excinfo:
Restore().run([
"get-basebackup",
"--config",
pghoard.config_path,
"--site",
pghoard.test_site,
"--target-dir",
backup_out,
"--tablespace-dir",
"tstest={}".format(backup_ts_out),
"--tablespace-dir",
"other={}".format(backup_other_out),
])
assert "Tablespace mapping for ['other'] was requested, but" in str(
excinfo.value)
# Now, finally, everything should be valid and we can proceed with restore
Restore().run([
"get-basebackup",
"--config",
pghoard.config_path,
"--site",
pghoard.test_site,
"--restore-to-master",
"--target-dir",
backup_out,
"--tablespace-dir",
"tstest={}".format(backup_ts_out),
])
# Adjust the generated recovery.conf to point pghoard_postgres_command to our instance
new_py_restore_cmd = "PYTHONPATH={} python3 -m pghoard.postgres_command --mode restore".format(
os.path.dirname(os.path.dirname(__file__)))
new_go_restore_cmd = "{}/pghoard_postgres_command_go --mode restore".format(
os.path.dirname(os.path.dirname(__file__)))
with open(os.path.join(backup_out, "recovery.conf"), "r+") as fp:
rconf = fp.read()
rconf = rconf.replace(
"pghoard_postgres_command_go --mode restore",
new_go_restore_cmd)
rconf = rconf.replace(
"pghoard_postgres_command --mode restore",
new_py_restore_cmd)
fp.seek(0)
fp.write(rconf)
r_db = PGTester(backup_out)
r_db.user = dict(db.user, host=backup_out)
r_db.run_pg()
r_conn_str = pgutil.create_connection_string(r_db.user)
# Wait for PG to start up
start_time = time.monotonic()
while True:
try:
r_conn = psycopg2.connect(r_conn_str)
break
except psycopg2.OperationalError as ex:
if "starting up" in str(ex):
assert time.monotonic() - start_time <= 10
time.sleep(1)
else:
raise
r_cursor = r_conn.cursor()
# Make sure the tablespace is defined and points to the right (new) path
r_cursor.execute(
"SELECT oid, pg_tablespace_location(oid) FROM pg_tablespace WHERE spcname = 'tstest'"
)
r_res = r_cursor.fetchone()
assert r_res[1] == backup_ts_out
# We should be able to read from the table in the tablespace and the values should match what we stored before
r_cursor.execute("SELECT id FROM tstest")
r_res = r_cursor.fetchall()
cursor.execute("SELECT id FROM tstest")
orig_res = cursor.fetchall()
assert r_res == orig_res
finally:
if r_conn:
r_conn.close()
if r_db:
r_db.kill(force=True)
cursor.execute("DROP TABLE IF EXISTS tstest")
cursor.execute("DROP TABLESPACE tstest")
conn.close()
def test_basebackups_tablespaces(self, capsys, db, pghoard, tmpdir):
# Create a test tablespace for this instance, but make sure we drop it at the end of the test as the
# database we use is shared by all test cases, and tablespaces are a global concept so the test
# tablespace could interfere with other tests
tspath = tmpdir.join("extra-ts").strpath
os.makedirs(tspath)
conn_str = pgutil.create_connection_string(db.user)
conn = psycopg2.connect(conn_str)
conn.autocommit = True
cursor = conn.cursor()
cursor.execute("CREATE TABLESPACE tstest LOCATION %s", [tspath])
r_db, r_conn = None, None
try:
cursor.execute("CREATE TABLE tstest (id BIGSERIAL PRIMARY KEY) TABLESPACE tstest")
cursor.execute("INSERT INTO tstest (id) VALUES (default)")
cursor.execute("SELECT oid, pg_tablespace_location(oid) FROM pg_tablespace WHERE spcname = 'tstest'")
res = cursor.fetchone()
assert res[1] == tspath
# Start receivexlog since we want the WALs to be able to restore later on
xlog_directory = os.path.join(pghoard.config["backup_location"], pghoard.test_site, "xlog_incoming")
makedirs(xlog_directory, exist_ok=True)
pghoard.receivexlog_listener(pghoard.test_site, db.user, xlog_directory)
cursor.execute("SELECT txid_current(), pg_switch_xlog()")
self._test_create_basebackup(capsys, db, pghoard, "local-tar")
cursor.execute("SELECT txid_current(), pg_switch_xlog()")
cursor.execute("SELECT txid_current(), pg_switch_xlog()")
backup_out = tmpdir.join("test-restore").strpath
backup_ts_out = tmpdir.join("test-restore-tstest").strpath
# Tablespaces are extracted to their previous absolute paths by default, but the path must be empty
# and it isn't as it's still used by the running PG
with pytest.raises(RestoreError) as excinfo:
Restore().run([
"get-basebackup",
"--config", pghoard.config_path,
"--site", pghoard.test_site,
"--target-dir", backup_out,
])
assert "Tablespace 'tstest' target directory" in str(excinfo.value)
assert "not empty" in str(excinfo.value)
# We can't restore tablespaces to non-existent directories either
with pytest.raises(RestoreError) as excinfo:
Restore().run([
"get-basebackup",
"--config", pghoard.config_path,
"--site", pghoard.test_site,
"--target-dir", backup_out,
"--tablespace-dir", "tstest={}".format(backup_ts_out),
])
assert "Tablespace 'tstest' target directory" in str(excinfo.value)
assert "does not exist" in str(excinfo.value)
os.makedirs(backup_ts_out)
# We can't restore if the directory isn't writable
os.chmod(backup_ts_out, 0o500)
with pytest.raises(RestoreError) as excinfo:
Restore().run([
"get-basebackup",
"--config", pghoard.config_path,
"--site", pghoard.test_site,
"--target-dir", backup_out,
"--tablespace-dir", "tstest={}".format(backup_ts_out),
])
assert "Tablespace 'tstest' target directory" in str(excinfo.value)
assert "empty, but not writable" in str(excinfo.value)
os.chmod(backup_ts_out, 0o700)
# We can't proceed if we request mappings for non-existent tablespaces
backup_other_out = tmpdir.join("test-restore-other").strpath
os.makedirs(backup_other_out)
with pytest.raises(RestoreError) as excinfo:
Restore().run([
"get-basebackup",
"--config", pghoard.config_path,
"--site", pghoard.test_site,
"--target-dir", backup_out,
"--tablespace-dir", "tstest={}".format(backup_ts_out),
"--tablespace-dir", "other={}".format(backup_other_out),
])
assert "Tablespace mapping for ['other'] was requested, but" in str(excinfo.value)
# Now, finally, everything should be valid and we can proceed with restore
Restore().run([
"get-basebackup",
"--config", pghoard.config_path,
"--site", pghoard.test_site,
"--restore-to-master",
"--target-dir", backup_out,
"--tablespace-dir", "tstest={}".format(backup_ts_out),
])
# Adjust the generated recovery.conf to point pghoard_postgres_command to our instance
new_cmd = "PYTHONPATH={} python3 -m pghoard.postgres_command".format(os.path.dirname(os.path.dirname(__file__)))
with open(os.path.join(backup_out, "recovery.conf"), "r+") as fp:
rconf = fp.read()
rconf = rconf.replace("pghoard_postgres_command", new_cmd)
fp.seek(0)
fp.write(rconf)
r_db = TestPG(backup_out)
r_db.user = dict(db.user, host=backup_out)
r_db.run_pg()
r_conn_str = pgutil.create_connection_string(r_db.user)
# Wait for PG to start up
start_time = time.monotonic()
while True:
try:
r_conn = psycopg2.connect(r_conn_str)
break
except psycopg2.OperationalError as ex:
if "starting up" in str(ex):
assert time.monotonic() - start_time <= 10
time.sleep(1)
else:
raise
r_cursor = r_conn.cursor()
# Make sure the tablespace is defined and points to the right (new) path
r_cursor.execute("SELECT oid, pg_tablespace_location(oid) FROM pg_tablespace WHERE spcname = 'tstest'")
r_res = r_cursor.fetchone()
assert r_res[1] == backup_ts_out
# We should be able to read from the table in the tablespace and the values should match what we stored before
r_cursor.execute("SELECT id FROM tstest")
r_res = r_cursor.fetchall()
cursor.execute("SELECT id FROM tstest")
orig_res = cursor.fetchall()
assert r_res == orig_res
finally:
if r_conn:
r_conn.close()
if r_db:
r_db.kill(force=True)
cursor.execute("DROP TABLE IF EXISTS tstest")
cursor.execute("DROP TABLESPACE tstest")
conn.close()
def _test_storage(st, driver, tmpdir, storage_config):
scratch = tmpdir.join("scratch")
compat.makedirs(str(scratch), exist_ok=True)
# File not found cases
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_metadata_for_key("NONEXISTENT")
with pytest.raises(errors.FileNotFoundFromStorageError):
st.delete_key("NONEXISTENT")
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_contents_to_file("NONEXISTENT", str(scratch.join("a")))
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_contents_to_fileobj("NONEXISTENT", BytesIO())
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_contents_to_string("NONEXISTENT")
assert st.list_path("") == []
assert st.list_path("NONEXISTENT") == []
st.store_file_from_memory("NONEXISTENT-a/x1", b"dummy", None)
dummy_file = str(scratch.join("a"))
with open(dummy_file, "wb") as fp:
fp.write(b"dummy")
st.store_file_from_disk("NONEXISTENT-b/x1", dummy_file, None)
st.store_file_from_disk("NONEXISTENT-b/x1", dummy_file, {"x": 1})
st.delete_key("NONEXISTENT-b/x1")
st.delete_key("NONEXISTENT-a/x1")
# Other basic cases
from_disk_file = str(scratch.join("a"))
with open(from_disk_file, "wb") as fp:
fp.write(b"from disk")
st.store_file_from_disk("test1/x1", from_disk_file, None)
out = BytesIO()
assert st.get_contents_to_fileobj("test1/x1", out) == {}
assert out.getvalue() == b"from disk"
if driver == "s3":
response = st.s3_client.head_object(
Bucket=st.bucket_name,
Key=st.format_key_for_backend("test1/x1"),
)
assert bool(response.get("ServerSideEncryption")) == bool(
storage_config.get('encrypted'))
st.store_file_from_memory("test1/x1", b"dummy", {"k": "v"})
out = BytesIO()
assert st.get_contents_to_fileobj("test1/x1", out) == {"k": "v"}
assert out.getvalue() == b"dummy"
st.store_file_from_memory("test1/x1", b"l", {"fancymetadata": "value"})
assert st.get_contents_to_string("test1/x1") == (b"l", {
"fancymetadata": "value"
})
st.store_file_from_memory("test1/x1", b"1", None)
assert st.get_contents_to_string("test1/x1") == (b"1", {})
st.store_file_from_memory("test1/td", b"to disk", {"to-disk": "42"})
to_disk_file = str(scratch.join("b"))
assert st.get_contents_to_file("test1/td", to_disk_file) == {
"to-disk": "42"
}
if driver == "s3":
response = st.s3_client.head_object(
Bucket=st.bucket_name,
Key=st.format_key_for_backend("test1/x1"),
)
assert bool(response.get("ServerSideEncryption")) == bool(
storage_config.get('encrypted'))
assert st.list_path("") == [
] # nothing at top level (directories not listed)
if driver == "local":
# create a dot-file (hidden), this must be ignored
target_file = os.path.join(st.prefix, "test1/.null")
with open(target_file, "w"):
pass
tlist = st.list_path("test1")
assert len(tlist) == 2
for fe in tlist:
assert isinstance(fe["last_modified"], datetime.datetime)
assert fe["last_modified"].tzinfo is not None
if fe["name"] == "test1/x1":
assert fe["size"] == 1
assert fe["metadata"] == {}
elif fe["name"] == "test1/td":
assert fe["size"] == len(b"to disk")
assert fe["metadata"] == {"to-disk": "42"}
else:
assert 0, "unexpected name in directory"
if driver == "google":
# test extra props for cacheControl in google
st.store_file_from_memory("test1/x1",
b"no cache test",
metadata={"test": "value"},
extra_props={"cacheControl": "no-cache"})
if driver == "local":
# test LocalFileIsRemoteFileError for local storage
target_file = os.path.join(st.prefix, "test1/x1")
with pytest.raises(errors.LocalFileIsRemoteFileError):
st.store_file_from_disk("test1/x1", target_file, {"local": True})
assert st.get_contents_to_string("test1/x1") == (b"1", {
"local": "True"
})
with pytest.raises(errors.LocalFileIsRemoteFileError):
st.get_contents_to_file("test1/x1", target_file)
# unlink metadata file, this shouldn't break anything
os.unlink(target_file + ".metadata")
assert st.get_metadata_for_key("test1/x1") == {}
st.delete_key("test1/x1")
st.delete_key("test1/td")
assert st.list_path("test1") == [] # empty again
test_hash = hashlib.sha256()
test_file = str(scratch.join("30m"))
test_size_send = 0
with open(test_file, "wb") as fp:
chunk = b"30m file" * 10000
while test_size_send < 30 * 1024 * 1024:
test_hash.update(chunk)
fp.write(chunk)
test_size_send += len(chunk)
test_hash_send = test_hash.hexdigest()
st.store_file_from_disk("test1/30m",
test_file,
multipart=True,
metadata={
"thirtymeg": "data",
"size": test_size_send,
"key": "value-with-a-hyphen"
})
os.unlink(test_file)
expected_meta = {
"thirtymeg": "data",
"size": str(test_size_send),
"key": "value-with-a-hyphen"
}
meta = st.get_metadata_for_key("test1/30m")
assert meta == expected_meta
progress_reports = []
def dl_progress(current_pos, expected_max):
progress_reports.append((current_pos, expected_max))
with open(test_file, "wb") as fp:
assert st.get_contents_to_fileobj(
"test1/30m", fp, progress_callback=dl_progress) == expected_meta
assert len(progress_reports) > 0
assert progress_reports[-1][0] == progress_reports[-1][1]
test_hash = hashlib.sha256()
test_size_rec = 0
with open(test_file, "rb") as fp:
while True:
chunk = fp.read(1024 * 1024)
if not chunk:
break
test_hash.update(chunk)
test_size_rec += len(chunk)
test_hash_rec = test_hash.hexdigest()
assert test_hash_rec == test_hash_send
assert test_size_rec == test_size_send
tlist = st.list_path("test1")
assert len(tlist) == 1
assert tlist[0]["name"] == "test1/30m"
assert tlist[0]["size"] == test_size_rec
if driver == "swift":
segments = test_size_send // st.segment_size
segment_list = st.list_path("test1_segments/30m")
assert len(segment_list) >= segments
if segments >= 2:
# reupload a file with the same name but with less chunks
os.truncate(test_file, st.segment_size + 1)
test_size_send = os.path.getsize(test_file)
st.store_file_from_disk("test1/30m",
test_file,
multipart=True,
metadata={
"30m": "less data",
"size": test_size_send
})
segment_list = st.list_path("test1_segments/30m")
assert len(segment_list) == 2
assert len(st.list_path("test1")) == 1
st.delete_key("test1/30m")
assert st.list_path("test1") == []
if driver == "swift":
assert st.list_path("test1_segments/30m") == []
def _test_storage(st, driver, tmpdir):
scratch = tmpdir.join("scratch")
compat.makedirs(str(scratch), exist_ok=True)
# File not found cases
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_metadata_for_key("NONEXISTENT")
with pytest.raises(errors.FileNotFoundFromStorageError):
st.delete_key("NONEXISTENT")
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_contents_to_file("NONEXISTENT", str(scratch.join("a")))
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_contents_to_fileobj("NONEXISTENT", BytesIO())
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_contents_to_string("NONEXISTENT")
assert st.list_path("") == []
assert st.list_path("NONEXISTENT") == []
st.store_file_from_memory("NONEXISTENT-a/x1", b"dummy", None)
dummy_file = str(scratch.join("a"))
with open(dummy_file, "wb") as fp:
fp.write(b"dummy")
st.store_file_from_disk("NONEXISTENT-b/x1", dummy_file, None)
st.store_file_from_disk("NONEXISTENT-b/x1", dummy_file, {"x": 1})
st.delete_key("NONEXISTENT-b/x1")
st.delete_key("NONEXISTENT-a/x1")
# Other basic cases
from_disk_file = str(scratch.join("a"))
with open(from_disk_file, "wb") as fp:
fp.write(b"from disk")
st.store_file_from_disk("test1/x1", from_disk_file, None)
out = BytesIO()
assert st.get_contents_to_fileobj("test1/x1", out) == {}
assert out.getvalue() == b"from disk"
st.store_file_from_memory("test1/x1", b"dummy", {"k": "v"})
out = BytesIO()
assert st.get_contents_to_fileobj("test1/x1", out) == {"k": "v"}
assert out.getvalue() == b"dummy"
st.store_file_from_memory("test1/x1", b"l", {"fancymetadata": "value"})
assert st.get_contents_to_string("test1/x1") == (b"l", {"fancymetadata": "value"})
st.store_file_from_memory("test1/x1", b"1", None)
assert st.get_contents_to_string("test1/x1") == (b"1", {})
st.store_file_from_memory("test1/td", b"to disk", {"to-disk": "42"})
to_disk_file = str(scratch.join("b"))
assert st.get_contents_to_file("test1/td", to_disk_file) == {"to-disk": "42"}
assert st.list_path("") == [] # nothing at top level (directories not listed)
if driver == "local":
# create a dot-file (hidden), this must be ignored
target_file = os.path.join(st.prefix, "test1/.null")
with open(target_file, "w"):
pass
tlist = st.list_path("test1")
assert len(tlist) == 2
for fe in tlist:
assert isinstance(fe["last_modified"], datetime.datetime)
assert fe["last_modified"].tzinfo is not None
if fe["name"] == "test1/x1":
assert fe["size"] == 1
assert fe["metadata"] == {}
elif fe["name"] == "test1/td":
assert fe["size"] == len(b"to disk")
assert fe["metadata"] == {"to-disk": "42"}
else:
assert 0, "unexpected name in directory"
if driver == "google":
# test extra props for cacheControl in google
st.store_file_from_memory("test1/x1", b"no cache test",
metadata={"test": "value"},
extra_props={"cacheControl": "no-cache"})
if driver == "local":
# test LocalFileIsRemoteFileError for local storage
target_file = os.path.join(st.prefix, "test1/x1")
with pytest.raises(errors.LocalFileIsRemoteFileError):
st.store_file_from_disk("test1/x1", target_file, {"local": True})
assert st.get_contents_to_string("test1/x1") == (b"1", {"local": "True"})
with pytest.raises(errors.LocalFileIsRemoteFileError):
st.get_contents_to_file("test1/x1", target_file)
# unlink metadata file, this shouldn't break anything
os.unlink(target_file + ".metadata")
assert st.get_metadata_for_key("test1/x1") == {}
st.delete_key("test1/x1")
st.delete_key("test1/td")
assert st.list_path("test1") == [] # empty again
test_hash = hashlib.sha256()
test_file = str(scratch.join("30m"))
test_size_send = 0
with open(test_file, "wb") as fp:
chunk = b"30m file" * 10000
while test_size_send < 30 * 1024 * 1024:
test_hash.update(chunk)
fp.write(chunk)
test_size_send += len(chunk)
test_hash_send = test_hash.hexdigest()
if driver == "s3":
# inject a failure in multipart uploads
def failing_new_key(key_name): # pylint: disable=unused-argument
# fail after the second call, restore functionality after the third
fail_calls[0] += 1
if fail_calls[0] > 3:
st.bucket.new_key = orig_new_key
if fail_calls[0] > 2:
raise Exception("multipart upload failure!")
fail_calls = [0]
orig_new_key = st.bucket.new_key
st.bucket.new_key = failing_new_key
st.store_file_from_disk("test1/30m", test_file, multipart=True,
metadata={"thirtymeg": "data", "size": test_size_send, "key": "value-with-a-hyphen"})
assert fail_calls[0] > 3
else:
st.store_file_from_disk("test1/30m", test_file, multipart=True,
metadata={"thirtymeg": "data", "size": test_size_send, "key": "value-with-a-hyphen"})
os.unlink(test_file)
expected_meta = {"thirtymeg": "data", "size": str(test_size_send), "key": "value-with-a-hyphen"}
meta = st.get_metadata_for_key("test1/30m")
assert meta == expected_meta
progress_reports = []
def dl_progress(current_pos, expected_max):
progress_reports.append((current_pos, expected_max))
with open(test_file, "wb") as fp:
assert st.get_contents_to_fileobj("test1/30m", fp, progress_callback=dl_progress) == expected_meta
assert len(progress_reports) > 0
assert progress_reports[-1][0] == progress_reports[-1][1]
test_hash = hashlib.sha256()
test_size_rec = 0
with open(test_file, "rb") as fp:
while True:
chunk = fp.read(1024 * 1024)
if not chunk:
break
test_hash.update(chunk)
test_size_rec += len(chunk)
test_hash_rec = test_hash.hexdigest()
assert test_hash_rec == test_hash_send
assert test_size_rec == test_size_send
tlist = st.list_path("test1")
assert len(tlist) == 1
assert tlist[0]["name"] == "test1/30m"
assert tlist[0]["size"] == test_size_rec
if driver == "swift":
segments = test_size_send // st.segment_size
segment_list = st.list_path("test1_segments/30m")
assert len(segment_list) >= segments
if segments >= 2:
# reupload a file with the same name but with less chunks
os.truncate(test_file, st.segment_size + 1)
test_size_send = os.path.getsize(test_file)
st.store_file_from_disk("test1/30m", test_file, multipart=True,
metadata={"30m": "less data", "size": test_size_send})
segment_list = st.list_path("test1_segments/30m")
assert len(segment_list) == 2
assert len(st.list_path("test1")) == 1
st.delete_key("test1/30m")
assert st.list_path("test1") == []
if driver == "swift":
assert st.list_path("test1_segments/30m") == []
def _test_storage(st, driver, tmpdir, storage_config):
scratch = tmpdir.join("scratch")
compat.makedirs(str(scratch), exist_ok=True)
# File not found cases
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_metadata_for_key("NONEXISTENT")
with pytest.raises(errors.FileNotFoundFromStorageError):
st.delete_key("NONEXISTENT")
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_contents_to_file("NONEXISTENT", str(scratch.join("a")))
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_contents_to_fileobj("NONEXISTENT", BytesIO())
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_contents_to_string("NONEXISTENT")
assert st.list_path("") == []
assert st.list_path("NONEXISTENT") == []
st.store_file_from_memory("NONEXISTENT-a/x1", b"dummy", None)
dummy_file = str(scratch.join("a"))
with open(dummy_file, "wb") as fp:
fp.write(b"dummy")
st.store_file_from_disk("NONEXISTENT-b/x1", dummy_file, None)
st.store_file_from_disk("NONEXISTENT-b/x1", dummy_file, {"x": 1})
st.delete_key("NONEXISTENT-b/x1")
st.delete_key("NONEXISTENT-a/x1")
# Other basic cases
from_disk_file = str(scratch.join("a"))
input_data = b"from disk"
if driver == "local":
input_data = input_data * 150000
with open(from_disk_file, "wb") as fp:
fp.write(input_data)
st.store_file_from_disk("test1/x1", from_disk_file, None)
out = BytesIO()
reported_positions = []
def progress_callback(pos, total):
reported_positions.append((pos, total))
assert st.get_contents_to_fileobj("test1/x1", out, progress_callback=progress_callback) == {}
assert out.getvalue() == input_data
if driver == "local":
input_size = len(input_data)
assert reported_positions == [(1024 * 1024, input_size), (input_size, input_size)]
if driver == "s3":
response = st.s3_client.head_object(
Bucket=st.bucket_name,
Key=st.format_key_for_backend("test1/x1"),
)
assert bool(response.get("ServerSideEncryption")) == bool(storage_config.get('encrypted'))
st.store_file_from_memory("test1/x1", b"dummy", {"k": "v"})
out = BytesIO()
assert st.get_contents_to_fileobj("test1/x1", out) == {"k": "v"}
assert out.getvalue() == b"dummy"
# Copy file
st.copy_file(source_key="test1/x1", destination_key="test_copy/copy1")
assert st.get_contents_to_string("test_copy/copy1") == (b"dummy", {"k": "v"})
st.copy_file(source_key="test1/x1", destination_key="test_copy/copy2", metadata={"new": "meta"})
assert st.get_contents_to_string("test_copy/copy2") == (b"dummy", {"new": "meta"})
st.store_file_from_memory("test1/x1", b"l", {"fancymetadata": "value"})
assert st.get_contents_to_string("test1/x1") == (b"l", {"fancymetadata": "value"})
st.store_file_from_memory("test1/x1", b"1", None)
assert st.get_contents_to_string("test1/x1") == (b"1", {})
st.store_file_from_memory("test1/td", b"to disk", {"to-disk": "42"})
to_disk_file = str(scratch.join("b"))
assert st.get_contents_to_file("test1/td", to_disk_file) == {"to-disk": "42"}
created_keys = {"test1/x1", "test1/td"}
if driver == "s3":
response = st.s3_client.head_object(
Bucket=st.bucket_name,
Key=st.format_key_for_backend("test1/x1"),
)
assert bool(response.get("ServerSideEncryption")) == bool(storage_config.get('encrypted'))
assert st.list_path("") == [] # nothing at top level (directories not listed)
if driver == "local":
# create a dot-file (hidden), this must be ignored
target_file = os.path.join(st.prefix, "test1/.null")
with open(target_file, "w"):
pass
tlist = st.list_path("test1")
assert len(tlist) == 2
for fe in tlist:
assert isinstance(fe["last_modified"], datetime.datetime)
assert fe["last_modified"].tzinfo is not None
if fe["name"] == "test1/x1":
assert fe["size"] == 1
assert fe["metadata"] == {}
elif fe["name"] == "test1/td":
assert fe["size"] == len(b"to disk")
assert fe["metadata"] == {"to-disk": "42"}
else:
assert 0, "unexpected name in directory"
assert set(st.iter_prefixes("test1")) == set()
for key in ["test1/sub1/sub1.1", "test1/sub2/sub2.1/sub2.1.1", "test1/sub3"]:
st.store_file_from_memory(key, b"1", None)
created_keys.add(key)
if driver == "local":
# sub3 is a file. Actual object storage systems support this, but a file system does not
with pytest.raises(NotADirectoryError):
st.store_file_from_memory("test1/sub3/sub3.1/sub3.1.1", b"1", None)
else:
st.store_file_from_memory("test1/sub3/sub3.1/sub3.1.1", b"1", None)
created_keys.add("test1/sub3/sub3.1/sub3.1.1")
if driver == "local":
assert set(st.iter_prefixes("test1")) == {"test1/sub1", "test1/sub2"}
else:
assert set(st.iter_prefixes("test1")) == {"test1/sub1", "test1/sub2", "test1/sub3"}
assert {item["name"] for item in st.list_path("test1")} == {"test1/x1", "test1/td", "test1/sub3"}
assert set(st.iter_prefixes("test1/sub1")) == set()
assert {item["name"] for item in st.list_path("test1/sub1")} == {"test1/sub1/sub1.1"}
assert {item["name"] for item in st.list_path("test1/sub2")} == set()
assert {item["name"] for item in st.list_path("test1/sub3")} == set()
assert set(st.iter_prefixes("test1/sub2")) == {"test1/sub2/sub2.1"}
if driver == "local":
assert set(st.iter_prefixes("test1/sub3")) == set() # sub3 is a file
else:
assert set(st.iter_prefixes("test1/sub3")) == {"test1/sub3/sub3.1"}
assert set(st.iter_prefixes("test1/sub3/3.1")) == set()
expected_deep_iter_test1_names = {
"test1/x1",
"test1/td",
"test1/sub1/sub1.1",
"test1/sub2/sub2.1/sub2.1.1",
"test1/sub3",
}
if driver != "local":
expected_deep_iter_test1_names.add("test1/sub3/sub3.1/sub3.1.1")
assert {item["name"] for item in st.list_path("test1", deep=True)} == expected_deep_iter_test1_names
def _object_names(iterable):
names = set()
for item in iterable:
assert item.type == KEY_TYPE_OBJECT
names.add(item.value["name"])
return names
deep_names_with_key = _object_names(st.iter_key("test1/sub3", deep=True, include_key=True))
deep_names_without_key = _object_names(st.iter_key("test1/sub3", deep=True, include_key=False))
if driver == "local":
assert deep_names_with_key == {"test1/sub3"}
assert deep_names_without_key == set()
else:
assert deep_names_with_key == {"test1/sub3", "test1/sub3/sub3.1/sub3.1.1"}
assert deep_names_without_key == {"test1/sub3/sub3.1/sub3.1.1"}
if driver == "google":
# test extra props for cacheControl in google
st.store_file_from_memory("test1/x1", b"no cache test",
metadata={"test": "value"},
extra_props={"cacheControl": "no-cache"})
if driver == "local":
# test LocalFileIsRemoteFileError for local storage
target_file = os.path.join(st.prefix, "test1/x1")
with pytest.raises(errors.LocalFileIsRemoteFileError):
st.store_file_from_disk("test1/x1", target_file, {"local": True})
assert st.get_contents_to_string("test1/x1") == (b"1", {"local": "True"})
with pytest.raises(errors.LocalFileIsRemoteFileError):
st.get_contents_to_file("test1/x1", target_file)
# Missing metadata is an error situation that should fail
os.unlink(target_file + ".metadata")
with pytest.raises(errors.FileNotFoundFromStorageError):
st.get_metadata_for_key("test1/x1")
for key in created_keys:
st.delete_key(key)
assert st.list_path("test1") == [] # empty again
for name in ["test2/foo", "test2/suba/foo", "test2/subb/bar", "test2/subb/subsub/zob"]:
st.store_file_from_memory(name, b"somedata")
names = sorted(item["name"] for item in st.list_path("test2", deep=True))
assert names == ["test2/foo", "test2/suba/foo", "test2/subb/bar", "test2/subb/subsub/zob"]
st.delete_tree("test2")
assert st.list_path("test2", deep=True) == []
test_hash = hashlib.sha256()
test_file = str(scratch.join("30m"))
test_size_send = 0
with open(test_file, "wb") as fp:
chunk = b"30m file" * 10000
while test_size_send < 30 * 1024 * 1024:
test_hash.update(chunk)
fp.write(chunk)
test_size_send += len(chunk)
test_hash_send = test_hash.hexdigest()
st.store_file_from_disk("test1/30m", test_file, multipart=True,
metadata={"thirtymeg": "data", "size": test_size_send, "key": "value-with-a-hyphen"})
os.unlink(test_file)
expected_meta = {"thirtymeg": "data", "size": str(test_size_send), "key": "value-with-a-hyphen"}
meta = st.get_metadata_for_key("test1/30m")
assert meta == expected_meta
progress_reports = []
def dl_progress(current_pos, expected_max):
progress_reports.append((current_pos, expected_max))
with open(test_file, "wb") as fp:
assert st.get_contents_to_fileobj("test1/30m", fp, progress_callback=dl_progress) == expected_meta
assert len(progress_reports) > 0
assert progress_reports[-1][0] == progress_reports[-1][1]
test_hash = hashlib.sha256()
test_size_rec = 0
with open(test_file, "rb") as fp:
while True:
chunk = fp.read(1024 * 1024)
if not chunk:
break
test_hash.update(chunk)
test_size_rec += len(chunk)
test_hash_rec = test_hash.hexdigest()
assert test_hash_rec == test_hash_send
assert test_size_rec == test_size_send
tlist = st.list_path("test1")
assert len(tlist) == 1
assert tlist[0]["name"] == "test1/30m"
assert tlist[0]["size"] == test_size_rec
if driver == "swift":
segments = test_size_send // st.segment_size
segment_list = st.list_path("test1_segments/30m")
assert len(segment_list) >= segments
if segments >= 2:
# reupload a file with the same name but with less chunks
os.truncate(test_file, st.segment_size + 1)
test_size_send = os.path.getsize(test_file)
st.store_file_from_disk("test1/30m", test_file, multipart=True,
metadata={"30m": "less data", "size": test_size_send})
segment_list = st.list_path("test1_segments/30m")
assert len(segment_list) == 2
assert len(st.list_path("test1")) == 1
st.delete_key("test1/30m")
assert st.list_path("test1") == []
if driver == "swift":
assert st.list_path("test1_segments/30m") == []
progress_reports = []
def upload_progress(progress):
progress_reports.append(progress)
for size in (300, 3 * 1024 * 1024, 11 * 1024 * 1024):
progress_reports = []
rds = RandomDataSource(size)
key = "test1/{}b".format(size)
st.store_file_object(key, rds, upload_progress_fn=upload_progress)
# Progress may be reported after each chunk and chunk size depends on available memory
# on current machine so there's no straightforward way of checking reasonable progress
# updates were made. Just ensure they're ordered correctly if something was provided
assert sorted(progress_reports) == progress_reports
bio = BytesIO()
st.get_contents_to_fileobj(key, bio)
buffer = bio.getbuffer()
assert len(buffer) == size
assert buffer == rds.data
st.delete_key(key)
