def run(self): try: properties_json = self.properties.value if properties_json.strip() != "": conf.override(conf.session_configs.__name__, json.loads(self.properties.value)) except ValueError as e: self.ipython_display.send_error("Session properties must be a valid JSON string. Error:\n{}".format(e)) return endpoint = self.endpoints_dropdown_widget.value language = self.lang_widget.value alias = self.session_widget.value skip = False properties = conf.get_session_properties(language) try: self.spark_controller.add_session(alias, endpoint, skip, properties) except ValueError as e: self.ipython_display.send_error("""Could not add session with name: {} properties: {} due to error: '{}'""".format(alias, properties, e)) return self.refresh_method()
def test_configuration_override(): kpc = { 'username': '******', 'password': '******', 'url': 'L' } overrides = { conf.kernel_python_credentials.__name__: kpc } conf.override_all(overrides) conf.override(conf.status_sleep_seconds.__name__, 1) assert_equals(conf._overrides, { conf.kernel_python_credentials.__name__: kpc, conf.status_sleep_seconds.__name__: 1 }) assert_equals(conf.status_sleep_seconds(), 1) assert_equals(conf.kernel_python_credentials(), kpc)
def spark(self, line, cell="", local_ns=None): """Magic to execute spark remotely. This magic allows you to create a Livy Scala or Python session against a Livy endpoint. Every session can be used to execute either Spark code or SparkSQL code by executing against the SQL context in the session. When the SQL context is used, the result will be a Pandas dataframe of a sample of the results. If invoked with no subcommand, the cell will be executed against the specified session. Subcommands ----------- info Display the available Livy sessions and other configurations for sessions. add Add a Livy session. First argument is the name of the session, second argument is the language, and third argument is the connection string of the Livy endpoint. A fourth argument specifying if session creation can be skipped if it already exists is optional: "skip" or empty. e.g. `%%spark add test python url=https://sparkcluster.net/livy;username=u;password=p skip` or e.g. `%%spark add test python url=https://sparkcluster.net/livy;username=u;password=p` config Override the livy session properties sent to Livy on session creation. All session creations will contain these config settings from then on. Expected value is a JSON key-value string to be sent as part of the Request Body for the POST /sessions endpoint in Livy. e.g. `%%spark config {"driverMemory":"1000M", "executorCores":4}` run Run Spark code against a session. e.g. `%%spark -s testsession` will execute the cell code against the testsession previously created e.g. `%%spark -s testsession -c sql` will execute the SQL code against the testsession previously created e.g. `%%spark -s testsession -c sql -o my_var` will execute the SQL code against the testsession previously created and store the pandas dataframe created in the my_var variable in the Python environment. logs Returns the logs for a given session. e.g. `%%spark logs -s testsession` will return the logs for the testsession previously created delete Delete a Livy session. Argument is the name of the session to be deleted. e.g. `%%spark delete defaultlivy` cleanup Delete all Livy sessions created by the notebook. No arguments required. e.g. `%%spark cleanup` """ usage = "Please look at usage of %spark by executing `%spark?`." user_input = line args = parse_argstring(self.spark, user_input) subcommand = args.command[0].lower() try: # info if subcommand == "info": if len(args.command) == 2: connection_string = args.command[1] info_sessions = self.spark_controller.get_all_sessions_endpoint_info(connection_string) self.print_endpoint_info(info_sessions) elif len(args.command) == 1: self._print_local_info() else: raise ValueError("Subcommand 'info' requires no value or a connection string to show all sessions.\n" "{}".format(usage)) # config elif subcommand == "config": # Would normally do " ".join(args.command[1:]) but parse_argstring removes quotes... rest_of_line = user_input[7:] conf.override(conf.session_configs.__name__, json.loads(rest_of_line)) # add elif subcommand == "add": if len(args.command) != 4 and len(args.command) != 5: raise ValueError("Subcommand 'add' requires three or four arguments.\n{}".format(usage)) name = args.command[1].lower() language = args.command[2].lower() connection_string = args.command[3] if len(args.command) == 5: skip = args.command[4].lower() == "skip" else: skip = False properties = conf.get_session_properties(language) self.spark_controller.add_session(name, connection_string, skip, properties) # delete elif subcommand == "delete": if len(args.command) == 2: name = args.command[1].lower() self.spark_controller.delete_session_by_name(name) elif len(args.command) == 3: connection_string = args.command[1] session_id = args.command[2] self.spark_controller.delete_session_by_id(connection_string, session_id) else: raise ValueError("Subcommand 'delete' requires a session name or a connection string and id.\n{}" .format(usage)) # cleanup elif subcommand == "cleanup": if len(args.command) == 2: connection_string = args.command[1] self.spark_controller.cleanup_endpoint(connection_string) elif len(args.command) == 1: self.spark_controller.cleanup() else: raise ValueError("Subcommand 'cleanup' requires no further values or a connection string to clean up " "sessions.\n{}".format(usage)) # logs elif subcommand == "logs": if len(args.command) == 1: (success, out) = self.spark_controller.get_logs(args.session) if success: self.ipython_display.write(out) else: self.ipython_display.send_error(out) else: raise ValueError("Subcommand 'logs' requires no further values.\n{}".format(usage)) # run elif len(subcommand) == 0: if args.context == CONTEXT_NAME_SPARK: (success, out) = self.spark_controller.run_cell(cell, args.session) if success: self.ipython_display.write(out) else: self.ipython_display.send_error(out) elif args.context == CONTEXT_NAME_SQL: sqlquery = SQLQuery(cell, args.samplemethod, args.maxrows, args.samplefraction) return self.execute_sqlquery(sqlquery, args.session, args.output, args.quiet) else: raise ValueError("Context '{}' not found".format(args.context)) # error else: raise ValueError("Subcommand '{}' not found. {}".format(subcommand, usage)) except ValueError as err: self.ipython_display.send_error("{}".format(err))
def _override_session_settings(settings): conf.override(conf.session_configs.__name__, settings)
def spark(self, line, cell="", local_ns=None): """Magic to execute spark remotely. This magic allows you to create a Livy Scala or Python session against a Livy endpoint. Every session can be used to execute either Spark code or SparkSQL code by executing against the SQL context in the session. When the SQL context is used, the result will be a Pandas dataframe of a sample of the results. If invoked with no subcommand, the cell will be executed against the specified session. Subcommands ----------- info Display the available Livy sessions and other configurations for sessions. add Add a Livy session given a session name (-s), language (-l), and endpoint credentials. The -k argument, if present, will skip adding this session if it already exists. e.g. `%spark add -s test -l python -u https://sparkcluster.net/livy -a u -p -k` config Override the livy session properties sent to Livy on session creation. All session creations will contain these config settings from then on. Expected value is a JSON key-value string to be sent as part of the Request Body for the POST /sessions endpoint in Livy. e.g. `%%spark config` `{"driverMemory":"1000M", "executorCores":4}` run Run Spark code against a session. e.g. `%%spark -s testsession` will execute the cell code against the testsession previously created e.g. `%%spark -s testsession -c sql` will execute the SQL code against the testsession previously created e.g. `%%spark -s testsession -c sql -o my_var` will execute the SQL code against the testsession previously created and store the pandas dataframe created in the my_var variable in the Python environment. logs Returns the logs for a given session. e.g. `%spark logs -s testsession` will return the logs for the testsession previously created delete Delete a Livy session. e.g. `%spark delete -s defaultlivy` cleanup Delete all Livy sessions created by the notebook. No arguments required. e.g. `%spark cleanup` """ usage = "Please look at usage of %spark by executing `%spark?`." user_input = line args = parse_argstring_or_throw(self.spark, user_input) subcommand = args.command[0].lower() # info if subcommand == "info": if args.url is not None: endpoint = Endpoint(args.url, args.user, args.password) info_sessions = self.spark_controller.get_all_sessions_endpoint_info(endpoint) self.print_endpoint_info(info_sessions) else: self._print_local_info() # config elif subcommand == "config": conf.override(conf.session_configs.__name__, json.loads(cell)) # add elif subcommand == "add": if args.url is None: self.ipython_display.send_error("Need to supply URL argument (e.g. -u https://example.com/livyendpoint)") return name = args.session language = args.language endpoint = Endpoint(args.url, args.user, args.password) skip = args.skip properties = conf.get_session_properties(language) self.spark_controller.add_session(name, endpoint, skip, properties) # delete elif subcommand == "delete": if args.session is not None: self.spark_controller.delete_session_by_name(args.session) elif args.url is not None: if args.id is None: self.ipython_display.send_error("Must provide --id or -i option to delete session at endpoint from URL") return endpoint = Endpoint(args.url, args.user, args.password) session_id = args.id self.spark_controller.delete_session_by_id(endpoint, session_id) else: self.ipython_display.send_error("Subcommand 'delete' requires a session name or a URL and session ID") # cleanup elif subcommand == "cleanup": if args.url is not None: endpoint = Endpoint(args.url, args.user, args.password) self.spark_controller.cleanup_endpoint(endpoint) else: self.spark_controller.cleanup() # logs elif subcommand == "logs": if args.session is None: self.ipython_display.send_error("Need to provide session argument (-s SESSION_NAME)") return self.ipython_display.write(self.spark_controller.get_logs(args.session)) # run elif len(subcommand) == 0: if args.session is None: self.ipython_display.send_error("Need to provide session argument (-s SESSION_NAME)") return if args.context == CONTEXT_NAME_SPARK: (success, out) = self.spark_controller.run_command(Command(cell), args.session) if success: self.ipython_display.write(out) else: self.ipython_display.send_error(out) elif args.context == CONTEXT_NAME_SQL: return self.execute_sqlquery(cell, args.samplemethod, args.maxrows, args.samplefraction, args.session, args.output, args.quiet) else: self.ipython_display.send_error("Context '{}' not found".format(args.context)) # error else: self.ipython_display.send_error("Subcommand '{}' not found. {}".format(subcommand, usage))
def test_configuration_raise_error_for_bad_base64_password(): kpc = { 'username': '******', 'base64_password': '******', 'url': 'L' } overrides = { conf.kernel_python_credentials.__name__: kpc } conf.override_all(overrides) conf.override(conf.status_sleep_seconds.__name__, 1) conf.kernel_python_credentials()
def _override_session_settings(settings): conf.override(conf.session_configs.__name__, json.loads(settings))