PT LAG and ECMP resolution require JSON-RPC ID number as input for later cancellation

Change-Id: I3d0139f88df1407ea1f5c1678055be2bb852038b

broadview_lib/config/agentapi.py

@@ -15,9 +15,12 @@
 import json
 from collections import OrderedDict
 from agentconnection import AgentConnection
+from broadviewconfig import BroadViewLibConfig
+import fcntl
 
 class AgentAPI(object):
-    __serial = 1
+    __serial = None
+    __cfg = BroadViewLibConfig()
     def __init__(self):
         self.__httpMethod = "POST"
         self.__feature = None
@@ -55,6 +58,88 @@ class AgentAPI(object):
         if self.__payload:
             ret = self.__payload["method"]
         return ret    
+    
+    def getLastUsedSerial(self):
+        '''
+        Some requests to the agent, like PT's get lag resolution with a
+        non-zero collection interval, require knowlege of the JSON RPC 
+        serial number because it is used as a key/index for later 
+        cancelling the collection. While this violates layering, this is 
+        no big deal if only one client (or client instance) ever connects 
+        to an agent in the data center, and the client never generates
+        (within reason) a request with a duplicate ID that maps to an ID
+        used by the agent as an index.
+         
+        When multiple clients, or even clients that are restarted enter 
+        the picture, requiring clients to generate IDs that are
+        unique across applications and hosts is obviously more difficult. 
+        Such difficulties are communication among clients within
+        the datacenter to ensure that no client uses the same IDs as any
+        other, and requiring each client to persist the last used ID in
+        case of crash or restart. 
+
+        The strategy we take here to deal with this is to use a file to 
+        store the last used ID, and protect that file with a SYSV file 
+        lock. If a client wants to make a request, the lock is acquired, 
+        the last used ID is read from the file, the ID is incremented by 
+        1 and the file rewritten, followed by release of the lock. Sharing 
+        across nodes in the datacenter can be acheived by all hosts using 
+        a lock file that is located in shared filesystem location, such 
+        as NFS. 
+
+        Obviously all apps on a host or within the datacenter that access
+        a given agent must use the same ID file. If there are multiple agents
+        in the datacenter (which is going to be the case, as no interesting
+        datacenter is going to have one switch), there must be one (and only
+        one) id file per switch/agent.
+
+        We can eliminate this need for request ID management once the
+        agent adopts a method where it allocates and returns handles for 
+        objects and tasks that it is managing to be used by clients to 
+        refer to those same objects and tasks in subsequent requests. 
+
+        This function exposes the last used ID so that the client can use
+        it in a subsequent (cancellation) request. 
+
+        
+        '''
+
+        return AgentAPI.__serial
+
+    def getIDFileLocation(self):
+        return AgentAPI.__cfg.getRequestIDFile()
+
+    def getNextSerial(self):
+
+        '''
+        Get the next request ID, updating the shared ID file. The name
+        of the ID file comes from a configuration file named 
+        /etc/broadviewlib.conf via a setting named json_rpc_id_path. Here 
+        is an example conf file that sets the ID file path to /tmp/foo.txt:
+
+        [misc]
+
+        json_rpc_id_path = /tmp/foo.txt
+
+        '''
+
+        filename = self.getIDFileLocation()
+
+        f = open(filename, "a").close()  # create if needed
+        f = open(filename, "r+")
+        fd = f.fileno()
+        fcntl.lockf(fd, fcntl.LOCK_EX)
+        serial = f.read()
+        if len(serial) == 0:
+            serial = 1        # new file, read zero bytes, so initialize
+        else:
+            serial = int(serial)
+        f.seek(0)
+        f.write("{}".format(serial + 1))
+        fcntl.lockf(fd, fcntl.LOCK_UN)
+        f.close()
+        AgentAPI.__serial = serial # XXX not thread safe
+        return serial
 
     def _send(self, o, timeout):
         self.__payload = {}
@@ -62,8 +147,7 @@ class AgentAPI(object):
         self.__payload["asic-id"] = o["asic-id"]
         self.__payload["method"] = o["method"]
         self.__payload["params"] = o["params"]
-        self.__payload["id"] = AgentAPI.__serial
+        self.__payload["id"] = self.getNextSerial()
-        AgentAPI.__serial = AgentAPI.__serial + 1
         conn = AgentConnection(self.__host, self.__port, self.__feature, timeout)
         r = conn.makeRequest(self)
         conn.close()

broadview_lib/config/broadviewconfig.py

@@ -16,6 +16,27 @@ import ConfigParser
 import json
 import unittest
 
+class BroadViewLibConfig():
+    cfg = None
+
+    def __init__(self):
+        if not BroadViewLibConfig.cfg:
+            try:
+                BroadViewLibConfig.cfg = ConfigParser.ConfigParser()
+                BroadViewLibConfig.cfg.read("/etc/broadviewlib.conf")
+            except:
+                BroadViewLibConfig.cfg = None
+                pass
+
+    def getRequestIDFile(self):
+        ret = "/tmp/bvserial.txt"
+        try:
+            ret = BroadViewLibConfig.cfg.get("misc", "json_rpc_id_path")
+        except:
+            pass
+
+        return ret
+
 class BroadViewBSTSwitches():
     cfg = None
     bst_switches = None

broadview_lib/config/pt.py

@@ -343,6 +343,8 @@ class GetPacketTraceLAGResolution(AgentAPI):
         rep = None
         if status == 200:
             self.__json = json["report"]
+            print self.__json
+            self.__json[0]["cancellation-id"] = self.getLastUsedSerial()
             rep = PTParser()
             rep.process(json)
         else:
@@ -402,6 +404,7 @@ class GetPacketTraceECMPResolution(AgentAPI):
         rep = None
         if status == 200:
             self.__json = json["report"]
+            self.__json[0]["cancellation-id"] = self.getLastUsedSerial()
             rep = PTParser()
             rep.process(json)
         else:
@@ -461,6 +464,7 @@ class GetPacketTraceProfile(AgentAPI):
         rep = None
         if status == 200:
             self.__json = json["report"]
+            self.__json[0]["cancellation-id"] = self.getLastUsedSerial()
             rep = PTParser()
             rep.process(json)
         else:

broadview_lib/tools/bv-ptctl.py

@@ -236,10 +236,24 @@ class PTCommand():
 
     def handleCancelProfile(self, args):
         usage = False
+	reqid = None
         usage, asic, host, port = self.getASICHostPort(args)
+        if not usage:
+            for arg in args:
+                if "id:" in arg:
+                    v = arg.split(":")
+                    if len(v) == 2:
+                        reqid = int(v[1])
+                    else:
+                        print "invalid id: bad argument count"
+                        usage = True
+        if reqid == None:
+            print "missing id argument"
+            usage = True
         if not usage:
             x = CancelPacketTraceProfile(host, port)
             x.setASIC(asic)
+            x.setId(reqid)
             status = x.send(self._timeout)
             if status != 200:
                 print "failure: {}".format(status)
@@ -249,13 +263,31 @@ class PTCommand():
 
     def helpCancelProfile(self, name):
         print name
+        print
+        print "args:"
+        print
+        print "   id:id (see cancellation-id output from get-profile command)"
+        print "   id:id"
 
     def handleCancelLAGResolution(self, args):
         usage = False
+	reqid = None
         usage, asic, host, port = self.getASICHostPort(args)
+        if not usage:
+            for arg in args:
+                if "id:" in arg:
+                    v = arg.split(":")
+                    if len(v) == 2:
+                        reqid = int(v[1])
+                    else:
+                        print "invalid id: bad argument count"
+                        usage = True
+        if reqid == None:
+            print "missing id argument"
         if not usage:
             x = CancelPacketTraceLAGResolution(host, port)
             x.setASIC(asic)
+            x.setId(reqid)
             status = x.send(self._timeout)
             if status != 200:
                 print "failure: {}".format(status)
@@ -265,13 +297,30 @@ class PTCommand():
 
     def helpCancelLAGResolution(self, name):
         print name
+        print
+        print "args:"
+        print
+        print "   id:id (see cancellation-id output from get-lag-resolution command"
 
     def handleCancelECMPResolution(self, args):
         usage = False
+	reqid = None
         usage, asic, host, port = self.getASICHostPort(args)
+        if not usage:
+            for arg in args:
+                if "id:" in arg:
+                    v = arg.split(":")
+                    if len(v) == 2:
+                        reqid = int(v[1])
+                    else:
+                        print "invalid id: bad argument count"
+                        usage = True
+        if reqid == None:
+            print "missing id argument"
         if not usage:
             x = CancelPacketTraceECMPResolution(host, port)
             x.setASIC(asic)
+            x.setId(reqid)
             status = x.send(self._timeout)
             if status != 200:
                 print "failure: {}".format(status)
@@ -281,6 +330,10 @@ class PTCommand():
 
     def helpCancelECMPResolution(self, name):
         print name
+        print
+        print "args:"
+        print
+        print "   id:id (see cancellation-id output from get-ecmp-resolution command"
 
     def handleCancelSendDropPacket(self, args):
         usage = False
@@ -450,6 +503,10 @@ class PTCommand():
         print "   port-list:port[,port][,port]...[,port]"
         print "   collection-interval: interval"
         print "   drop-packet: [0|1]"
+        print
+        print "notes:"
+        print
+        print "   use the cancellation-id output as id arg for cancel-lag-resolution command to cancel when collection-interval non-zero"
 
     def handleGetECMPResolution(self, args):
         usage = False
@@ -510,6 +567,10 @@ class PTCommand():
         print "   port-list:port[,port][,port]...[,port]"
         print "   collection-interval: interval"
         print "   drop-packet: [0|1]"
+        print
+        print "notes:"
+        print
+        print "   use the cancellation-id output as id arg for cancel-ecmp-resolution command to cancel when collection-interval non-zero"
 
     def handleGetProfile(self, args):
         usage = False
@@ -575,6 +636,10 @@ class PTCommand():
         print "   port-list:port[,port][,port]...[,port]"
         print "   collection-interval: interval"
         print "   drop-packet: [0|1]"
+        print
+        print "notes:"
+        print
+        print "   use the cancellation-id output as id arg for cancel-profile command to cancel when collection-interval non-zero"
 
     def handleGetDropReason(self, args):
         usage = False