X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/231db3a5eb4b9876acfe780ba0404e73a30b302f..3f1e065d5095b2c0cda036a130575458c8f270af:/lib/luxi.py diff --git a/lib/luxi.py b/lib/luxi.py index ad11aef..cfde21e 100644 --- a/lib/luxi.py +++ b/lib/luxi.py @@ -1,7 +1,7 @@ # # -# Copyright (C) 2006, 2007 Google Inc. +# Copyright (C) 2006, 2007, 2011 Google Inc. # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by @@ -34,16 +34,20 @@ import collections import time import errno import logging +import warnings from ganeti import serializer from ganeti import constants from ganeti import errors +from ganeti import utils +from ganeti import objects KEY_METHOD = "method" KEY_ARGS = "args" KEY_SUCCESS = "success" KEY_RESULT = "result" +KEY_VERSION = "version" REQ_SUBMIT_JOB = "SubmitJob" REQ_SUBMIT_MANY_JOBS = "SubmitManyJobs" @@ -51,33 +55,41 @@ REQ_WAIT_FOR_JOB_CHANGE = "WaitForJobChange" REQ_CANCEL_JOB = "CancelJob" REQ_ARCHIVE_JOB = "ArchiveJob" REQ_AUTOARCHIVE_JOBS = "AutoArchiveJobs" +REQ_QUERY = "Query" +REQ_QUERY_FIELDS = "QueryFields" REQ_QUERY_JOBS = "QueryJobs" REQ_QUERY_INSTANCES = "QueryInstances" REQ_QUERY_NODES = "QueryNodes" +REQ_QUERY_GROUPS = "QueryGroups" REQ_QUERY_EXPORTS = "QueryExports" REQ_QUERY_CONFIG_VALUES = "QueryConfigValues" REQ_QUERY_CLUSTER_INFO = "QueryClusterInfo" +REQ_QUERY_TAGS = "QueryTags" +REQ_QUERY_LOCKS = "QueryLocks" REQ_QUEUE_SET_DRAIN_FLAG = "SetDrainFlag" REQ_SET_WATCHER_PAUSE = "SetWatcherPause" DEF_CTMO = 10 DEF_RWTO = 60 +# WaitForJobChange timeout +WFJC_TIMEOUT = (DEF_RWTO - 1) / 2 -class ProtocolError(errors.GenericError): - """Denotes an error in the LUXI protocol""" + +class ProtocolError(errors.LuxiError): + """Denotes an error in the LUXI protocol.""" class ConnectionClosedError(ProtocolError): - """Connection closed error""" + """Connection closed error.""" class TimeoutError(ProtocolError): - """Operation timeout error""" + """Operation timeout error.""" class RequestError(ProtocolError): - """Error on request + """Error on request. This signifies an error in the request format or request handling, but not (e.g.) an error in starting up an instance. @@ -90,7 +102,7 @@ class RequestError(ProtocolError): class NoMasterError(ProtocolError): - """The master cannot be reached + """The master cannot be reached. This means that the master daemon is not running or the socket has been removed. @@ -98,6 +110,14 @@ class NoMasterError(ProtocolError): """ +class PermissionError(ProtocolError): + """Permission denied while connecting to the master socket. + + This means the user doesn't have the proper rights. + + """ + + class Transport: """Low-level transport class. @@ -110,15 +130,12 @@ class Transport: """ - def __init__(self, address, timeouts=None, eom=None): + def __init__(self, address, timeouts=None): """Constructor for the Client class. Arguments: - address: a valid address the the used transport class - timeout: a list of timeouts, to be used on connect and read/write - - eom: an identifier to be used as end-of-message which the - upper-layer will guarantee that this identifier will not appear - in any message There are two timeouts used since we might want to wait for a long time for a response, but the connect timeout should be lower. @@ -141,22 +158,16 @@ class Transport: self._buffer = "" self._msgs = collections.deque() - if eom is None: - self.eom = '\3' - else: - self.eom = eom - try: self.socket = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM) - self.socket.settimeout(self._ctimeout) + + # Try to connect try: - self.socket.connect(address) - except socket.timeout, err: - raise TimeoutError("Connect timed out: %s" % str(err)) - except socket.error, err: - if err.args[0] in (errno.ENOENT, errno.ECONNREFUSED): - raise NoMasterError(address) - raise + utils.Retry(self._Connect, 1.0, self._ctimeout, + args=(self.socket, address, self._ctimeout)) + except utils.RetryTimeout: + raise TimeoutError("Connect timed out") + self.socket.settimeout(self._rwtimeout) except (socket.error, NoMasterError): if self.socket is not None: @@ -164,6 +175,24 @@ class Transport: self.socket = None raise + @staticmethod + def _Connect(sock, address, timeout): + sock.settimeout(timeout) + try: + sock.connect(address) + except socket.timeout, err: + raise TimeoutError("Connect timed out: %s" % str(err)) + except socket.error, err: + error_code = err.args[0] + if error_code in (errno.ENOENT, errno.ECONNREFUSED): + raise NoMasterError(address) + elif error_code in (errno.EPERM, errno.EACCES): + raise PermissionError(address) + elif error_code == errno.EAGAIN: + # Server's socket backlog is full at the moment + raise utils.RetryAgain() + raise + def _CheckSocket(self): """Make sure we are connected. @@ -177,13 +206,13 @@ class Transport: This just sends a message and doesn't wait for the response. """ - if self.eom in msg: + if constants.LUXI_EOM in msg: raise ProtocolError("Message terminator found in payload") self._CheckSocket() try: # TODO: sendall is not guaranteed to send everything - self.socket.sendall(msg + self.eom) + self.socket.sendall(msg + constants.LUXI_EOM) except socket.timeout, err: raise TimeoutError("Sending timeout: %s" % str(err)) @@ -204,16 +233,16 @@ class Transport: while True: try: data = self.socket.recv(4096) + except socket.timeout, err: + raise TimeoutError("Receive timeout: %s" % str(err)) except socket.error, err: if err.args and err.args[0] == errno.EAGAIN: continue raise - except socket.timeout, err: - raise TimeoutError("Receive timeout: %s" % str(err)) break if not data: raise ConnectionClosedError("Connection closed while reading") - new_msgs = (self._buffer + data).split(self.eom) + new_msgs = (self._buffer + data).split(constants.LUXI_EOM) self._buffer = new_msgs.pop() self._msgs.extend(new_msgs) return self._msgs.popleft() @@ -249,14 +278,16 @@ def ParseRequest(msg): logging.error("LUXI request not a dict: %r", msg) raise ProtocolError("Invalid LUXI request (not a dict)") - method = request.get(KEY_METHOD, None) - args = request.get(KEY_ARGS, None) + method = request.get(KEY_METHOD, None) # pylint: disable-msg=E1103 + args = request.get(KEY_ARGS, None) # pylint: disable-msg=E1103 + version = request.get(KEY_VERSION, None) # pylint: disable-msg=E1103 + if method is None or args is None: logging.error("LUXI request missing method or arguments: %r", msg) raise ProtocolError(("Invalid LUXI request (no method or arguments" " in request): %r") % msg) - return (method, args) + return (method, args, version) def ParseResponse(msg): @@ -266,6 +297,8 @@ def ParseResponse(msg): # Parse the result try: data = serializer.LoadJson(msg) + except KeyboardInterrupt: + raise except Exception, err: raise ProtocolError("Error while deserializing response: %s" % str(err)) @@ -275,10 +308,11 @@ def ParseResponse(msg): KEY_RESULT in data): raise ProtocolError("Invalid response from server: %r" % data) - return (data[KEY_SUCCESS], data[KEY_RESULT]) + return (data[KEY_SUCCESS], data[KEY_RESULT], + data.get(KEY_VERSION, None)) # pylint: disable-msg=E1103 -def FormatResponse(success, result): +def FormatResponse(success, result, version=None): """Formats a LUXI response message. """ @@ -287,12 +321,15 @@ def FormatResponse(success, result): KEY_RESULT: result, } + if version is not None: + response[KEY_VERSION] = version + logging.debug("LUXI response: %s", response) return serializer.DumpJson(response) -def FormatRequest(method, args): +def FormatRequest(method, args, version=None): """Formats a LUXI request message. """ @@ -302,22 +339,30 @@ def FormatRequest(method, args): KEY_ARGS: args, } + if version is not None: + request[KEY_VERSION] = version + # Serialize the request return serializer.DumpJson(request, indent=False) -def CallLuxiMethod(transport_cb, method, args): +def CallLuxiMethod(transport_cb, method, args, version=None): """Send a LUXI request via a transport and return the response. """ assert callable(transport_cb) - request_msg = FormatRequest(method, args) + request_msg = FormatRequest(method, args, version=version) # Send request and wait for response response_msg = transport_cb(request_msg) - (success, result) = ParseResponse(response_msg) + (success, result, resp_version) = ParseResponse(response_msg) + + # Verify version if there was one in the response + if resp_version is not None and resp_version != version: + raise errors.LuxiError("LUXI version mismatch, client %s, response %s" % + (version, resp_version)) if success: return result @@ -384,11 +429,18 @@ class Client(object): self._CloseTransport() raise + def Close(self): + """Close the underlying connection. + + """ + self._CloseTransport() + def CallMethod(self, method, args): """Send a generic request and return the response. """ - return CallLuxiMethod(self._SendMethodCall, method, args) + return CallLuxiMethod(self._SendMethodCall, method, args, + version=constants.LUXI_VERSION) def SetQueueDrainFlag(self, drain_flag): return self.CallMethod(REQ_QUEUE_SET_DRAIN_FLAG, drain_flag) @@ -416,16 +468,65 @@ class Client(object): timeout = (DEF_RWTO - 1) / 2 return self.CallMethod(REQ_AUTOARCHIVE_JOBS, (age, timeout)) + def WaitForJobChangeOnce(self, job_id, fields, + prev_job_info, prev_log_serial, + timeout=WFJC_TIMEOUT): + """Waits for changes on a job. + + @param job_id: Job ID + @type fields: list + @param fields: List of field names to be observed + @type prev_job_info: None or list + @param prev_job_info: Previously received job information + @type prev_log_serial: None or int/long + @param prev_log_serial: Highest log serial number previously received + @type timeout: int/float + @param timeout: Timeout in seconds (values larger than L{WFJC_TIMEOUT} will + be capped to that value) + + """ + assert timeout >= 0, "Timeout can not be negative" + return self.CallMethod(REQ_WAIT_FOR_JOB_CHANGE, + (job_id, fields, prev_job_info, + prev_log_serial, + min(WFJC_TIMEOUT, timeout))) + def WaitForJobChange(self, job_id, fields, prev_job_info, prev_log_serial): - timeout = (DEF_RWTO - 1) / 2 while True: - result = self.CallMethod(REQ_WAIT_FOR_JOB_CHANGE, - (job_id, fields, prev_job_info, - prev_log_serial, timeout)) + result = self.WaitForJobChangeOnce(job_id, fields, + prev_job_info, prev_log_serial) if result != constants.JOB_NOTCHANGED: break return result + def Query(self, what, fields, filter_): + """Query for resources/items. + + @param what: One of L{constants.QR_VIA_LUXI} + @type fields: List of strings + @param fields: List of requested fields + @type filter_: None or list + @param filter_: Query filter + @rtype: L{objects.QueryResponse} + + """ + req = objects.QueryRequest(what=what, fields=fields, filter=filter_) + result = self.CallMethod(REQ_QUERY, req.ToDict()) + return objects.QueryResponse.FromDict(result) + + def QueryFields(self, what, fields): + """Query for available fields. + + @param what: One of L{constants.QR_VIA_LUXI} + @type fields: None or list of strings + @param fields: List of requested fields + @rtype: L{objects.QueryFieldsResponse} + + """ + req = objects.QueryFieldsRequest(what=what, fields=fields) + result = self.CallMethod(REQ_QUERY_FIELDS, req.ToDict()) + return objects.QueryFieldsResponse.FromDict(result) + def QueryJobs(self, job_ids, fields): return self.CallMethod(REQ_QUERY_JOBS, (job_ids, fields)) @@ -435,6 +536,9 @@ class Client(object): def QueryNodes(self, names, fields, use_locking): return self.CallMethod(REQ_QUERY_NODES, (names, fields, use_locking)) + def QueryGroups(self, names, fields, use_locking): + return self.CallMethod(REQ_QUERY_GROUPS, (names, fields, use_locking)) + def QueryExports(self, nodes, use_locking): return self.CallMethod(REQ_QUERY_EXPORTS, (nodes, use_locking)) @@ -444,5 +548,10 @@ class Client(object): def QueryConfigValues(self, fields): return self.CallMethod(REQ_QUERY_CONFIG_VALUES, fields) + def QueryTags(self, kind, name): + return self.CallMethod(REQ_QUERY_TAGS, (kind, name)) -# TODO: class Server(object) + def QueryLocks(self, fields, sync): + warnings.warn("This LUXI call is deprecated and will be removed, use" + " Query(\"%s\", ...) instead" % constants.QR_LOCK) + return self.CallMethod(REQ_QUERY_LOCKS, (fields, sync))