def __init__(
self,
config: PbenchServerConfig,
logger: Logger,
schema: Schema,
*,
role: API_OPERATION = API_OPERATION.READ,
):
"""
Base class constructor.
Args:
config: server configuration
logger: logger object
schema: API schema: for example,
Schema(
Parameter("user", ParamType.USER, required=True),
Parameter("start", ParamType.DATE)
)
role: specify the API role, defaulting to READ
NOTE: each class currently only supports one HTTP method, so we can
describe only one set of parameters. If we ever need to change this,
we can add a level and describe distinct parameters for each method.
"""
super().__init__()
self.logger = logger
self.prefix = config.get("Indexing", "index_prefix")
host = config.get("elasticsearch", "host")
port = config.get("elasticsearch", "port")
self.es_url = f"http://{host}:{port}"
self.schema = schema
self.role = role
def main(cfg_name):
if not cfg_name:
print(
f"{_NAME_}: ERROR: No config file specified; set"
" _PBENCH_SERVER_CONFIG env variable or use --config <file> on the"
" command line",
file=sys.stderr,
)
return 2
try:
config = PbenchServerConfig(cfg_name)
except BadConfig as e:
print(f"{_NAME_}: {e} (config file {cfg_name})", file=sys.stderr)
return 1
logger = get_pbench_logger(_NAME_, config)
qdir, receive_dir = fetch_config_val(config, logger)
if qdir is None and receive_dir is None:
return 2
qdir_md5 = qdirs_check("quarantine", Path(qdir, "md5-002"), logger)
duplicates = qdirs_check("duplicates", Path(qdir, "duplicates-002"),
logger)
# The following directory holds tarballs that are quarantined because
# of operational errors on the server. They should be retried after
# the problem is fixed: basically, move them back into the reception
# area for 002 agents and wait.
errors = qdirs_check("errors", Path(qdir, "errors-002"), logger)
if qdir_md5 is None or duplicates is None or errors is None:
return 1
counts = process_tb(config, logger, receive_dir, qdir_md5, duplicates,
errors)
result_string = (f"{config.TS}: Processed {counts.ntotal} entries,"
f" {counts.ntbs} tarballs successful,"
f" {counts.nquarantined} quarantined tarballs,"
f" {counts.ndups} duplicately-named tarballs,"
f" {counts.nerrs} errors.")
logger.info(result_string)
# prepare and send report
with tempfile.NamedTemporaryFile(mode="w+t", dir=config.TMP) as reportfp:
reportfp.write(f"{counts.nstatus}{result_string}\n")
reportfp.seek(0)
report = Report(config, _NAME_)
report.init_report_template()
try:
report.post_status(config.timestamp(), "status", reportfp.name)
except Exception as exc:
logger.warning("Report post Unsuccesful: '{}'", exc)
return 0
def test_pbench_logger_level():
config = PbenchServerConfig(cfg_name)
logger = get_pbench_logger(_NAME_, config)
logging_level = config.get("logging", "logging_level")
logger = mock_the_handler(logger, logging_level, log_files[logging_level])
logger.debug(log_msgs["10"])
logger.info(log_msgs["20"])
logger.warning(log_msgs["30"])
logger.error(log_msgs["40"])
logger.critical(log_msgs["50"])
def test_pbench_logger():
config = PbenchServerConfig(cfg_name)
logger = get_pbench_logger(_NAME_, config)
logger_type = config.get("logging", "logger_type")
logger = mock_the_handler(logger, logger_type, log_files[logger_type])
logger.debug(log_msgs[logger_type])
if os.path.isfile(os.path.join(logdir, log_files[logger_type])):
with open(os.path.join(logdir, log_files[logger_type]), "r") as f:
assert (
f.read()[:-1] == log_msgs[logger_type]
), "Mismatch: the file did not contain the expected message."
def create_app():
"""Create Flask app with defined resource endpoints."""
global app
cfg_name = os.environ.get("_PBENCH_SERVER_CONFIG")
if not cfg_name:
print(
f"{__name__}: ERROR: No config file specified; set"
" _PBENCH_SERVER_CONFIG",
file=sys.stderr,
)
sys.exit(1)
try:
config = PbenchServerConfig(cfg_name)
except BadConfig as e:
print(f"{__name__}: {e} (config file {cfg_name})", file=sys.stderr)
sys.exit(1)
app = Flask(__name__)
api = Api(app)
app.logger = get_pbench_logger(__name__, config)
app.config_server = config.conf["pbench-server"]
app.config_elasticsearch = config.conf["elasticsearch"]
app.config_graphql = config.conf["graphql"]
prdp = app.config_server.get("pbench-receive-dir-prefix")
if not prdp:
app.logger.error(
"Missing config variable for pbench-receive-dir-prefix")
sys.exit(1)
try:
upload_directory = Path(f"{prdp}-002").resolve(strict=True)
except FileNotFoundError:
app.logger.exception(
"pbench-receive-dir-prefix does not exist on the host")
sys.exit(1)
except Exception:
app.logger.exception(
"Exception occurred during setting up the upload directory on the host"
)
sys.exit(1)
else:
app.upload_directory = upload_directory
app.config["PORT"] = app.config_server.get("rest_port")
app.config["VERSION"] = app.config_server.get("rest_version")
app.config["MAX_CONTENT_LENGTH"] = filesize_bytes(
app.config_server.get("rest_max_content_length"))
app.config["REST_URI"] = app.config_server.get("rest_uri")
app.config["LOG"] = app.config_server.get("rest_log")
app.config["BIND_HOST"] = app.config_server.get("bind_host")
app.config["WORKERS"] = app.config_server.get("workers")
register_endpoints(api, app)
return app
def __init__(self, config: PbenchServerConfig, logger: Logger):
"""
__init__ Construct the API resource
Args:
:config: server config values
:logger: message logging
Report the server configuration to a web client. By default, the Pbench
server ansible script sets up a local Apache reverse proxy routing
through the HTTP port (80); an external reverse-proxy can be configured
without the knowledge of the server, and this API will use reverse-proxy
Forwarded or X-Forwarded-Host HTTP headers to discover the proxy
configuration. All server endpoints will be reported with respect to that
address.
"""
self.logger = logger
self.host = config.get("pbench-server", "host")
self.uri_prefix = config.rest_uri
self.prefix = config.get("Indexing", "index_prefix")
self.commit_id = config.COMMIT_ID
def get_server_config():
cfg_name = os.environ.get("_PBENCH_SERVER_CONFIG")
if not cfg_name:
raise ConfigFileNotSpecified(
f"{__name__}: ERROR: No config file specified; set"
" _PBENCH_SERVER_CONFIG")
try:
return PbenchServerConfig(cfg_name)
except BadConfig as e:
raise Exception(
f"{__name__}: {e} (config file {cfg_name})").with_traceback(
e.__traceback__)
def __init__(self, config: PbenchServerConfig, logger: Logger, schema: Schema):
"""
__init__ Construct the base class
Args:
config: server configuration
logger: logger object
schema: API schema: for example,
Schema(
Parameter("user", ParamType.USER, required=True),
Parameter("start", ParamType.DATE)
)
NOTE: each class currently only supports one HTTP method, so we can
describe only one set of parameters. If we ever need to change this,
we can add a level and describe parameters by method.
"""
super().__init__()
self.logger = logger
self.prefix = config.get("Indexing", "index_prefix")
host = config.get("elasticsearch", "host")
port = config.get("elasticsearch", "port")
self.es_url = f"http://{host}:{port}"
self.schema = schema
def main(options):
if not options.cfg_name:
print(
f"{_NAME_}: ERROR: No config file specified; set"
" _PBENCH_SERVER_CONFIG env variable",
file=sys.stderr,
)
return 1
try:
config = PbenchServerConfig(options.cfg_name)
except BadConfig as e:
print(f"{_NAME_}: {e}", file=sys.stderr)
return 2
logger = get_pbench_logger(_NAME_, config)
# We're going to need the Postgres DB to track dataset state, so setup
# DB access.
init_db(config, logger)
# NOTE: the importer will ignore datasets that already exist in the DB;
# we do the ACTION links first to get set up, and then sweep other links
# to record the state of remaining datasets, especially to record those
# which are quarantined.
#
# FIXME: This doesn't sweep the "<root>/quarantine" directory, which might
# have additional datasets. Are they worth importing?
actions = {
"TO-UNPACK": States.UPLOADED,
"TO-INDEX": States.UNPACKED,
"INDEXED": States.INDEXED,
"UNPACKED": States.UNPACKED,
"WONT-UNPACK": States.QUARANTINED,
"WONT-INDEX*": States.QUARANTINED,
"BAD-MD5": States.QUARANTINED,
}
importer = Import(logger, options, config)
return_value = 0
for link, state in actions.items():
status = importer.process(link, state)
if status != 0:
return_value = 1
return return_value
def create_app():
global app
cfg_name = os.environ.get("_PBENCH_SERVER_CONFIG")
if not cfg_name:
print(
f"{__name__}: ERROR: No config file specified; set"
" _PBENCH_SERVER_CONFIG",
file=sys.stderr,
)
sys.exit(1)
try:
config = PbenchServerConfig(cfg_name)
except BadConfig as e:
print(f"{__name__}: {e} (config file {cfg_name})", file=sys.stderr)
sys.exit(1)
app = Flask(__name__)
api = Api(app)
app.logger = get_pbench_logger(__name__, config)
app.config_server = config.conf["pbench-server"]
prdp = app.config_server.get("pbench-receive-dir-prefix")
try:
upload_directory = Path(f"{prdp}-002").resolve(strict=True)
except Exception:
app.logger.error(
"Missing config variable for pbench-receive-dir-prefix")
sys.exit(1)
else:
app.upload_directory = upload_directory
app.config["PORT"] = app.config_server.get("rest_port")
app.config["VERSION"] = app.config_server.get("rest_version")
app.config["MAX_CONTENT_LENGTH"] = app.config_server.get(
"rest_max_content_length")
app.config["REST_URI"] = app.config_server.get("rest_uri")
app.config["LOG"] = app.config_server.get("rest_log")
register_endpoints(api, app)
return app
def execute(self):
config = PbenchServerConfig(self.context.config)
logger = get_pbench_logger(_NAME_, config)
# We're going to need the Postgres DB to track dataset state, so setup
# DB access.
Database.init_db(config, logger)
user = User(
username=self.context.username,
password=self.context.password,
first_name=self.context.first_name,
last_name=self.context.last_name,
email=self.context.email,
role=self.context.role if self.context.role else "",
)
user.add()
if user.is_admin():
click.echo(f"Admin user {self.context.username} registered")
else:
click.echo(f"User {self.context.username} registered")
help="The caller's user ID (optional)",
)
parser.add_argument(
"-T",
"--type",
dest="doctype",
required=True,
help="The type of report document to index, one of status|error",
)
parser.add_argument(
"file_to_index", nargs=1, help="The file containing the report to index"
)
parsed = parser.parse_args()
try:
config = PbenchServerConfig(parsed.cfg_name)
except BadConfig as e:
print(f"{_prog}: {e}", file=sys.stderr)
sys.exit(1)
# We're going to need the Postgres DB to track dataset state, so setup
# DB access. We don't pass a Logger here, because that introduces lots
# of spurious changes in the gold CLI test output.
init_db(config, None)
hostname = gethostname()
pid = parsed.pid
group_id = parsed.group_id
user_id = parsed.user_id
report = Report(
# environment variable set. Since we really need access to the config
# file to operate, and we know the relative location of that config file,
# we check to see if that exists before declaring a problem.
config_path = Path(_dir.parent, "lib", "config", "pbench-server.cfg")
if not config_path.exists():
print(
f"{_prog}: No config file specified: set _PBENCH_SERVER_CONFIG env variable or use"
f" --config <file> on the command line",
file=sys.stderr,
)
sys.exit(1)
else:
config_path = Path(parsed.cfg_name)
try:
config = PbenchServerConfig(config_path)
except BadConfig as e:
print(f"{_prog}: {e} (config file {config_path})", file=sys.stderr)
sys.exit(1)
# Exclude the "files" and "conf" attributes from being exported
vars = sorted([
key for key in config.__dict__.keys()
if key not in ("files", "conf", "timestamp", "_unittests", "_ref_datetime",
"get")
])
for att in vars:
try:
os.environ[att] = str(getattr(config, att))
except AttributeError:
print(
def main(cfg_name):
if not cfg_name:
print(
f"{_NAME_}: ERROR: No config file specified; set"
" _PBENCH_SERVER_CONFIG env variable or use --config <file> on the"
" command line",
file=sys.stderr,
)
return 2
try:
config = PbenchServerConfig(cfg_name)
except BadConfig as e:
print(f"{_NAME_}: {e}", file=sys.stderr)
return 1
logger = get_pbench_logger(_NAME_, config)
# We're going to need the Postgres DB to track dataset state, so setup
# DB access.
init_db(config, logger)
# Add a BACKUP and QDIR field to the config object
config.BACKUP = config.conf.get("pbench-server", "pbench-backup-dir")
config.QDIR = config.get("pbench-server", "pbench-quarantine-dir")
# call the LocalBackupObject class
lb_obj = LocalBackupObject(config)
# call the S3Config class
s3_obj = S3Config(config, logger)
lb_obj, s3_obj = sanity_check(lb_obj, s3_obj, config, logger)
if lb_obj is None and s3_obj is None:
return 3
logger.info("start-{}", config.TS)
# Initiate the backup
counts = backup_data(lb_obj, s3_obj, config, logger)
result_string = (f"Total processed: {counts.ntotal},"
f" Local backup successes: {counts.nbackup_success},"
f" Local backup failures: {counts.nbackup_fail},"
f" S3 upload successes: {counts.ns3_success},"
f" S3 upload failures: {counts.ns3_fail},"
f" Quarantined: {counts.nquaran}")
logger.info(result_string)
prog = Path(sys.argv[0]).name
# prepare and send report
with tempfile.NamedTemporaryFile(mode="w+t", dir=config.TMP) as reportfp:
reportfp.write(
f"{prog}.{config.timestamp()}({config.PBENCH_ENV})\n{result_string}\n"
)
reportfp.seek(0)
report = Report(config, _NAME_)
report.init_report_template()
try:
report.post_status(config.timestamp(), "status", reportfp.name)
except Exception:
pass
logger.info("end-{}", config.TS)
return 0
# file to operate, and we know the relative location of that config file,
# we check to see if that exists before declaring a problem.
config_name = os.path.join(os.path.dirname(_dir), "lib", "config",
"pbench-server.cfg")
if not os.path.exists(config_name):
print(
f"{_prog}: No config file specified: set _PBENCH_SERVER_CONFIG env variable or use"
f" --config <file> on the command line",
file=sys.stderr,
)
sys.exit(1)
else:
config_name = parsed.cfg_name
try:
config = PbenchServerConfig(config_name)
except BadConfig as e:
print(f"{_prog}: {e} (config file {config_name})", file=sys.stderr)
sys.exit(1)
# Exclude the "files" and "conf" attributes from being exported
vars = sorted([
key for key in config.__dict__.keys()
if key not in ("files", "conf", "timestamp", "_unittests", "_ref_datetime",
"get")
])
for att in vars:
try:
os.environ[att] = str(getattr(config, att))
except AttributeError:
print(
def main(options):
if not options.cfg_name:
print(
f"{_NAME_}: ERROR: No config file specified; set"
" _PBENCH_SERVER_CONFIG env variable",
file=sys.stderr,
)
return 1
try:
config = PbenchServerConfig(options.cfg_name)
except BadConfig as e:
print(f"{_NAME_}: {e}", file=sys.stderr)
return 2
try:
archive_p = Path(config.ARCHIVE).resolve(strict=True)
except FileNotFoundError:
print(
f"The configured ARCHIVE directory, {config.ARCHIVE}, does not exist",
file=sys.stderr,
)
return 3
if not archive_p.is_dir():
print(
f"The configured ARCHIVE directory, {config.ARCHIVE}, is not a valid directory",
file=sys.stderr,
)
return 4
try:
incoming_p = Path(config.INCOMING).resolve(strict=True)
except FileNotFoundError:
print(
f"The configured INCOMING directory, {config.INCOMING}, does not exist",
file=sys.stderr,
)
return 5
if not incoming_p.is_dir():
print(
f"The configured INCOMING directory, {config.INCOMING}, is not a valid directory",
file=sys.stderr,
)
return 6
_fmt = "%Y-%m-%d"
try:
oldest_dt = datetime.strptime(options.oldest, _fmt)
newest_dt = datetime.strptime(options.newest, _fmt)
except Exception as exc:
print(
f"Invalid time range, {options.oldest} to {options.newest}, "
f"'{exc}', expected time range values in the form YYYY-MM-DD",
file=sys.stderr,
)
return 7
else:
if newest_dt < oldest_dt:
# For convenience, swap oldest and newest dates that are reversed.
oldest_dt, newest_dt = newest_dt, oldest_dt
print(f"Re-indexing tar balls in the range {oldest_dt} to {newest_dt}")
actions = []
start = pbench.server._time()
for _val in gen_reindex_list(archive_p, oldest_dt, newest_dt):
controller_name, tb_name = _val
act_set = reindex(
controller_name, tb_name, archive_p, incoming_p, options.dry_run
)
actions.append(act_set)
end = pbench.server._time()
for act_set in sorted(actions):
print(f"{act_set!r}")
print(f"Run-time: {start} {end} {end - start}")
return 0
def main(options):
try:
if not options.cfg_name:
print(
f"{_NAME_}: ERROR: No config file specified; set"
" _PBENCH_SERVER_CONFIG env variable",
file=sys.stderr,
)
return 1
try:
config = PbenchServerConfig(options.cfg_name)
except BadConfig as e:
print(f"{_NAME_}: {e}", file=sys.stderr)
return 2
logger = get_pbench_logger(_NAME_, config)
# We're going to need the Postgres DB to track dataset state, so setup
# DB access.
init_db(config, logger)
args = {}
if options.create:
args["owner"] = options.create
if options.controller:
args["controller"] = options.controller
if options.path:
args["path"] = options.path
if options.name:
args["name"] = options.name
if options.md5:
args["md5"] = options.md5
if options.state:
try:
new_state = States[options.state.upper()]
except KeyError:
print(
f"{_NAME_}: Specified string '{options.state}' is not a Pbench dataset state",
file=sys.stderr,
)
return 1
args["state"] = new_state
if "path" not in args and ("controller" not in args or "name" not in args):
print(
f"{_NAME_}: Either --path or both --controller and --name must be specified",
file=sys.stderr,
)
return 1
# Either create a new dataset or attach to an existing dataset
doit = Dataset.create if options.create else Dataset.attach
# Find or create the specified dataset.
doit(**args)
except Exception as e:
# Stringify any exception and report it; then fail
logger.exception("Failed")
print(f"{_NAME_}: {e}", file=sys.stderr)
return 1
else:
return 0
def main(options):
if not options.cfg_name:
print(
f"{_NAME_}: ERROR: No config file specified; set"
" _PBENCH_SERVER_CONFIG env variable",
file=sys.stderr,
)
return 1
try:
config = PbenchServerConfig(options.cfg_name)
except BadConfig as e:
print(f"{_NAME_}: {e}", file=sys.stderr)
return 2
logger = get_pbench_logger(_NAME_, config)
archivepath = config.ARCHIVE
incoming = config.INCOMING
incomingpath = config.get_valid_dir_option("INCOMING", incoming, logger)
if not incomingpath:
return 3
results = config.RESULTS
resultspath = config.get_valid_dir_option("RESULTS", results, logger)
if not resultspath:
return 3
users = config.USERS
userspath = config.get_valid_dir_option("USERS", users, logger)
if not userspath:
return 3
# Fetch the configured maximum number of days a tar can remain "unpacked"
# in the INCOMING tree.
try:
max_unpacked_age = config.conf.get("pbench-server", "max-unpacked-age")
except NoOptionError as e:
logger.error(f"{e}")
return 5
try:
max_unpacked_age = int(max_unpacked_age)
except Exception:
logger.error("Bad maximum unpacked age, {}", max_unpacked_age)
return 6
# First phase is to find all the tar balls which are beyond the max
# unpacked age, and which still have an unpacked directory in INCOMING.
if config._ref_datetime is not None:
try:
curr_dt = config._ref_datetime
except Exception:
# Ignore bad dates from test environment.
curr_dt = datetime.utcnow()
else:
curr_dt = datetime.utcnow()
_msg = "Culling unpacked tar balls {} days older than {}"
if options.dry_run:
print(
_msg.format(max_unpacked_age, curr_dt.strftime(_STD_DATETIME_FMT)))
else:
logger.debug(_msg, max_unpacked_age,
curr_dt.strftime(_STD_DATETIME_FMT))
actions_taken = []
errors = 0
start = pbench.server._time()
gen = gen_list_unpacked_aged(incomingpath, archivepath, curr_dt,
max_unpacked_age)
if config._unittests:
# force the generator and sort the list
gen = sorted(list(gen))
for tb_incoming_dir, controller_name in gen:
act_set = remove_unpacked(
tb_incoming_dir,
controller_name,
resultspath,
userspath,
logger,
options.dry_run,
)
unpacked_dir_name = Path(tb_incoming_dir).name
act_path = Path(controller_name, unpacked_dir_name)
act_set.set_name(act_path)
actions_taken.append(act_set)
if act_set.errors > 0:
# Stop any further unpacked tar ball removal if an error is
# encountered.
break
end = pbench.server._time()
# Generate the ${TOP}/public_html prefix so we can strip it from the
# various targets in the report.
public_html = os.path.realpath(os.path.join(config.TOP, "public_html"))
# Write the actions taken into a report file.
with tempfile.NamedTemporaryFile(mode="w+t",
prefix=f"{_NAME_}.",
suffix=".report",
dir=config.TMP) as tfp:
duration = end - start
total = len(actions_taken)
print(
f"Culled {total:d} unpacked tar ball directories ({errors:d}"
f" errors) in {duration:0.2f} secs",
file=tfp,
)
if total > 0:
print("\nActions Taken:", file=tfp)
for act_set in sorted(actions_taken, key=lambda a: a.name):
print(
f" - {act_set.name} ({act_set.errors:d} errors,"
f" {act_set.duration():0.2f} secs)",
file=tfp,
)
for act in act_set.actions:
assert act.noun.startswith(
public_html
), f"Logic bomb! {act.noun} not in .../public_html/"
tgt = Path(act.noun[len(public_html) + 1:])
if act.verb == "mv":
name = tgt.name
controller = tgt.parent
ex_tgt = controller / f".delete.{name}"
print(f" $ {act.verb} {tgt} {ex_tgt} # {act.status}",
file=tfp)
else:
print(f" $ {act.verb} {tgt} # {act.status}",
file=tfp)
# Flush out the report ahead of posting it.
tfp.flush()
tfp.seek(0)
# We need to generate a report that lists all the actions taken.
report = Report(config, _NAME_)
report.init_report_template()
try:
report.post_status(config.timestamp(),
"status" if errors == 0 else "errors", tfp.name)
except Exception:
pass
return errors
def main(options, name):
"""Main entry point to pbench-index.
The caller is required to pass the "options" argument with the following
expected attributes:
cfg_name - Name of the configuration file to use
dump_index_patterns - Don't do any indexing, but just emit the
list of index patterns that would be used
dump_templates - Dump the templates that would be used
index_tool_data - Index tool data only
re_index - Consider tar balls marked for re-indexing
All exceptions are caught and logged to syslog with the stacktrace of
the exception in a sub-object of the logged JSON document.
Signal Handlers used to establish different patterns for the three
behaviors:
1. Gracefully stop processing tar balls
- SIGQUIT
- The current tar ball is indexed until completion, but no other
tar balls are processed.
- Handler Behavior:
- Sets a flag that causes the code flow to break out of the
for loop.
- Does not raise an exception.
2. Interrupt the current tar ball being indexed, and proceed to the
next one, if any
- SIGINT
- Handler Behavior:
- try/except/finally placed immediately around the es_index()
call so that the signal handler will only be established for
the duration of the call.
- Raises an exception caught by above try/except/finally.
- The finally clause would take down the signal handler.
3. Stop processing tar balls immediately and exit gracefully
- SIGTERM
- Handler Behavior:
- Raises an exception caught be a new, outer-most, try/except
block that does not have a finally clause (as you don't want
any code execution in the finally block).
4. Re-evaluate the list of tar balls to index after indexing of the
current tar ball has finished
- SIGHUP
- Report the current state of indexing
- Current tar ball being processed
- Count of Remaining tarballs
- Count of Completed tarballs
- No. of Errors encountered
- Handler Behavior:
- No exception raised
"""
_name_suf = "-tool-data" if options.index_tool_data else ""
_name_re = "-re" if options.re_index else ""
name = f"{name}{_name_re}{_name_suf}"
error_code = Index.error_code
if not options.cfg_name:
print(
f"{name}: ERROR: No config file specified; set"
" _PBENCH_SERVER_CONFIG env variable or"
" use --config <file> on the command line",
file=sys.stderr,
)
return error_code["CFG_ERROR"].value
idxctx = None
try:
# We're going to need the Postgres DB to track dataset state, so setup
# DB access. We need to do this before we create the IdxContext, since
# that will need the template DB; so we create a PbenchServerConfig and
# Logger independently.
config = PbenchServerConfig(options.cfg_name)
logger = get_pbench_logger(name, config)
init_db(config, logger)
# Now we can initialize the index context
idxctx = IdxContext(options, name, config, logger, _dbg=_DEBUG)
except (ConfigFileError, ConfigParserError) as e:
print(f"{name}: {e}", file=sys.stderr)
return error_code["CFG_ERROR"].value
except BadConfig as e:
print(f"{name}: {e}", file=sys.stderr)
return error_code["BAD_CFG"].value
except JsonFileError as e:
print(f"{name}: {e}", file=sys.stderr)
return error_code["MAPPING_ERROR"].value
if options.dump_index_patterns:
idxctx.templates.dump_idx_patterns()
return 0
if options.dump_templates:
idxctx.templates.dump_templates()
return 0
res = error_code["OK"]
ARCHIVE_rp = idxctx.config.ARCHIVE
INCOMING_rp = idxctx.config.INCOMING
INCOMING_path = idxctx.config.get_valid_dir_option("INCOMING", INCOMING_rp,
idxctx.logger)
if not INCOMING_path:
res = error_code["BAD_CFG"]
qdir = idxctx.config.get_conf("QUARANTINE", "pbench-server",
"pbench-quarantine-dir", idxctx.logger)
if not qdir:
res = error_code["BAD_CFG"]
else:
qdir_path = idxctx.config.get_valid_dir_option("QDIR", Path(qdir),
idxctx.logger)
if not qdir_path:
res = error_code["BAD_CFG"]
if not res.success:
# Exit early if we encounter any errors.
return res.value
idxctx.logger.debug("{}.{}: starting", name, idxctx.TS)
index_obj = Index(name, options, idxctx, INCOMING_rp, ARCHIVE_rp, qdir)
status, tarballs = index_obj.collect_tb()
if status == 0 and tarballs:
status = index_obj.process_tb(tarballs)
return status
def main():
cfg_name = os.environ.get("_PBENCH_SERVER_CONFIG")
if not cfg_name:
print(
f"{_NAME_}: ERROR: No config file specified; set _PBENCH_SERVER_CONFIG env variable or"
f" use --config <file> on the command line",
file=sys.stderr,
)
return 2
try:
config = PbenchServerConfig(cfg_name)
except BadConfig as e:
print(f"{_NAME_}: {e}", file=sys.stderr)
return 1
logger = get_pbench_logger(_NAME_, config)
# We're going to need the Postgres DB to track dataset state, so setup
# DB access.
init_db(config, logger)
# add a BACKUP field to the config object
config.BACKUP = backup = config.conf.get("pbench-server",
"pbench-backup-dir")
if not backup:
logger.error(
"Unspecified backup directory, no pbench-backup-dir config in"
" pbench-server section")
return 1
backuppath = config.get_valid_dir_option("BACKUP", backup, logger)
if not backuppath:
return 1
# instantiate the s3config class
s3_config_obj = S3Config(config, logger)
s3_config_obj = sanity_check(s3_config_obj, logger)
logger.info("start-{}", config.TS)
start = config.timestamp()
prog = Path(sys.argv[0]).name
sts = 0
# N.B. tmpdir is the pathname of the temp directory.
with tempfile.TemporaryDirectory() as tmpdir:
archive_obj = BackupObject("ARCHIVE", config.ARCHIVE, tmpdir, logger)
local_backup_obj = BackupObject("BACKUP", config.BACKUP, tmpdir,
logger)
s3_backup_obj = BackupObject("S3", s3_config_obj, tmpdir, logger)
with tempfile.NamedTemporaryFile(mode="w+t", dir=tmpdir) as reportfp:
reportfp.write(
f"{prog}.{config.TS} ({config.PBENCH_ENV}) started at {start}\n"
)
if s3_config_obj is None:
reportfp.write(
"\nNOTICE: S3 backup service is inaccessible; skipping"
" ARCHIVE to S3 comparison\n\n")
# FIXME: Parallelize these three ...
# Create entry list for archive
logger.debug("Starting archive list creation")
ar_start = config.timestamp()
ret_sts = archive_obj.entry_list_creation()
if ret_sts == Status.FAIL:
sts += 1
logger.debug("Finished archive list ({!r})", ret_sts)
# Create entry list for backup
logger.debug("Starting local backup list creation")
lb_start = config.timestamp()
ret_sts = local_backup_obj.entry_list_creation()
if ret_sts == Status.FAIL:
sts += 1
logger.debug("Finished local backup list ({!r})", ret_sts)
# Create entry list for S3
if s3_config_obj is not None:
logger.debug("Starting S3 list creation")
s3_start = config.timestamp()
ret_sts = s3_backup_obj.entry_list_creation()
if ret_sts == Status.FAIL:
sts += 1
logger.debug("Finished S3 list ({!r})", ret_sts)
logger.debug("Checking MD5 signatures of archive")
ar_md5_start = config.timestamp()
try:
# Check the data integrity in ARCHIVE (Question 1).
md5_result_archive = archive_obj.checkmd5()
except Exception as ex:
msg = f"Failed to check data integrity of ARCHIVE ({config.ARCHIVE})"
logger.exception(msg)
reportfp.write(f"\n{msg} - '{ex}'\n")
sts += 1
else:
if md5_result_archive > 0:
# Create a report for failed MD5 results from ARCHIVE (Question 1)
archive_obj.report_failed_md5(reportfp)
sts += 1
logger.debug(
"Checking MD5 signature of archive: {} errors",
md5_result_archive,
)
logger.debug("Finished checking MD5 signatures of archive")
logger.debug("Checking MD5 signatures of local backup")
lb_md5_start = config.timestamp()
try:
# Check the data integrity in BACKUP (Question 2).
md5_result_backup = local_backup_obj.checkmd5()
except Exception as ex:
msg = f"Failed to check data integrity of BACKUP ({config.BACKUP})"
logger.exception(msg)
reportfp.write(f"\n{msg} - '{ex}'\n")
else:
if md5_result_backup > 0:
# Create a report for failed MD5 results from BACKUP (Question 2)
local_backup_obj.report_failed_md5(reportfp)
sts += 1
logger.debug(
"Checking MD5 signature of local backup: {} errors",
md5_result_backup,
)
logger.debug("Finished checking MD5 signatures of local backup")
# Compare ARCHIVE with BACKUP (Questions 3 and 3a).
msg = "Comparing ARCHIVE with BACKUP"
reportfp.write(f"\n{msg}\n{'-' * len(msg)}\n")
logger.debug("{}: start", msg)
compare_entry_lists(archive_obj, local_backup_obj, reportfp,
logger)
logger.debug("{}: end", msg)
if s3_config_obj is not None:
# Compare ARCHIVE with S3 (Questions 4, 4a, and 4b).
msg = "Comparing ARCHIVE with S3"
reportfp.write(f"\n{msg}\n{'-' * len(msg)}\n")
logger.debug("{}: start", msg)
compare_entry_lists(archive_obj, s3_backup_obj, reportfp,
logger)
logger.debug("{}: end", msg)
if s3_config_obj is None:
s3_start = "<skipped>"
reportfp.write(f"\n\nPhases (started):\n"
f"Archive List Creation: {ar_start}\n"
f"Local Backup List Creation: {lb_start}\n"
f"S3 List Creation: {s3_start}\n"
f"Archive MD5 Checks: {ar_md5_start}\n"
f"Local Backup MD5 Checks: {lb_md5_start}\n")
end = config.timestamp()
reportfp.write(
f"\n{prog}.{config.TS} ({config.PBENCH_ENV}) finished at {end}\n"
)
# Rewind to the beginning.
reportfp.seek(0)
logger.debug("Sending report: start")
report = Report(config, _NAME_)
report.init_report_template()
try:
report.post_status(config.timestamp(), "status", reportfp.name)
except Exception:
pass
logger.debug("Sending report: end")
logger.info("end-{}", config.TS)
return sts
def config_setup(context: object) -> None:
config = PbenchServerConfig(context.config)
# We're going to need the Postgres DB to track dataset state, so setup
# DB access.
init_db(config, None)
def main(options):
try:
if not options.cfg_name:
print(
f"{_NAME_}: ERROR: No config file specified; set"
" _PBENCH_SERVER_CONFIG env variable",
file=sys.stderr,
)
return 1
try:
config = PbenchServerConfig(options.cfg_name)
except BadConfig as e:
print(f"{_NAME_}: {e}", file=sys.stderr)
return 2
logger = get_pbench_logger(_NAME_, config)
# We're going to need the Postgres DB to track dataset state, so setup
# DB access.
Database.init_db(config, logger)
args = {}
if options.create:
user = options.create
try:
user = Auth.validate_user(user)
except UnknownUser:
# FIXME: I don't want to be creating the user here or
# dealing with a non-existing user. The unittest setup
# should create the test users we want ahead of time
# using a pbench-user-manager command and we should
# depend on having them here! The following is a hack
# until that command exists.
#
# The desired behavior would be to remove this try and
# except and allow UnknownUser to be handled below with
# an error message and termination.
User(
username=user,
first_name=user.capitalize(),
last_name="Account",
password=f"{user}_password",
email=f"{user}@example.com",
).add()
args["owner"] = user
if options.controller:
args["controller"] = options.controller
if options.path:
args["path"] = options.path
if options.name:
args["name"] = options.name
if options.md5:
args["md5"] = options.md5
if options.state:
try:
new_state = States[options.state.upper()]
except KeyError:
print(
f"{_NAME_}: Specified string '{options.state}' is not a Pbench dataset state",
file=sys.stderr,
)
return 1
args["state"] = new_state
if "path" not in args and ("controller" not in args
or "name" not in args):
print(
f"{_NAME_}: Either --path or both --controller and --name must be specified",
file=sys.stderr,
)
return 1
# Either create a new dataset or attach to an existing dataset
doit = Dataset.create if options.create else Dataset.attach
# Find or create the specified dataset.
doit(**args)
except Exception as e:
# Stringify any exception and report it; then fail
logger.exception("Failed")
print(f"{_NAME_}: {e}", file=sys.stderr)
return 1
else:
return 0
def main(options):
if not options.tb_path:
print(
f"{_NAME_}: ERROR: No tar ball path specified",
file=sys.stderr,
)
return 2
try:
tb_path = Path(options.tb_path).resolve(strict=True)
except FileNotFoundError:
print(
f"The Tarball Path, '{options.tb_path}', does not resolve to a real location"
)
else:
tb_name = tb_path.name
if not options.cfg_name:
print(
f"{_NAME_}: ERROR: No config file specified; set"
" _PBENCH_SERVER_CONFIG env variable",
file=sys.stderr,
)
return 3
try:
config = PbenchServerConfig(options.cfg_name)
except BadConfig as e:
print(f"{_NAME_}: {e}", file=sys.stderr)
return 4
archivepath = config.ARCHIVE
incoming = config.INCOMING
try:
incomingpath = incoming.resolve(strict=True)
except FileNotFoundError:
print(
f"The configured INCOMING directory, {incoming}, does not exist",
file=sys.stderr,
)
return 7
else:
if not incomingpath.is_dir():
print(
f"The configured INCOMING directory, {incoming},"
" is not a valid directory",
file=sys.stderr,
)
return 8
# Fetch the configured maximum number of days a tar can remain "unpacked"
# in the INCOMING tree.
try:
max_unpacked_age = config.conf.get("pbench-server", "max-unpacked-age")
except NoOptionError as e:
print(f"{e}", file=sys.stderr)
return 9
try:
max_unpacked_age = int(max_unpacked_age)
except Exception:
print(f"Bad maximum unpacked age, {max_unpacked_age}", file=sys.stderr)
return 10
# Check the unpacked directory name pattern.
match = tb_pat.fullmatch(tb_name)
if not match:
print(f"Unrecognized tar ball name format, {tb_name}", file=sys.stderr)
return 11
if not str(tb_path).startswith(str(archivepath)):
print(f"Given tar ball, {tb_path}, not from the ARCHIVE tree",
file=sys.stderr)
return 12
if not tb_path.exists():
print(
f"Given tar ball, {tb_path}, does not seem to exist in the ARCHIVE tree",
file=sys.stderr,
)
return 13
# Determine the proper time to use as a reference.
if config._ref_datetime is not None:
try:
curr_dt = config._ref_datetime
except Exception:
# Ignore bad dates from test environment.
curr_dt = datetime.utcnow()
else:
curr_dt = datetime.utcnow()
# Turn the pattern components of the match into a datetime object.
tb_dt = datetime(
int(match.group(1)),
int(match.group(2)),
int(match.group(3)),
int(match.group(4)),
int(match.group(5)),
int(match.group(6)),
)
# See if this unpacked tar ball directory has "aged" out.
timediff = curr_dt - tb_dt
if timediff.days > max_unpacked_age:
# Finally, make one last check to see if this tar ball
# directory should be kept regardless of aging out.
controller_p = tb_path.parent.name
if Path(incomingpath, controller_p, tb_name,
".__pbench_keep__").is_file():
ret_val = 0
else:
ret_val = 1
else:
ret_val = 0
return ret_val