cleanup slice-fs-* steps, nicer output, and shows contents when it's wrong
[tests.git] / system / TestSlice.py
index be8ad58..f5049d3 100644 (file)
@@ -1,12 +1,40 @@
+# Thierry Parmentelat <thierry.parmentelat@inria.fr>
+# Copyright (C) 2010 INRIA 
+#
 import utils
 import os, os.path
-import datetime
+from datetime import datetime, timedelta
 import time
 
 from TestKey import TestKey
 from TestUser import TestUser
-from TestNode import TestNode
+from TestNode import TestNode, CompleterTaskNodeSsh
 from TestSsh import TestSsh
+from Completer import Completer, CompleterTask
+
+class CompleterTaskSliceSsh (CompleterTask):
+
+    def __init__ (self, test_plc, hostname, slicename, private_key,command, expected, dry_run):
+        self.test_plc=test_plc
+        self.hostname=hostname
+        self.slicename=slicename
+        self.private_key=private_key
+        self.command=command
+        self.dry_run=dry_run
+        self.expected=expected
+    def run (self, silent): 
+        (site_spec,node_spec) = self.test_plc.locate_hostname(self.hostname)
+        test_ssh = TestSsh (self.hostname,key=self.private_key,username=self.slicename)
+        full_command = test_ssh.actual_command(self.command)
+        retcod = utils.system (full_command, silent=silent)
+        if self.dry_run: return True
+        if self.expected:       return retcod==0
+        else:                   return retcod!=0
+    def failure_message (self):
+        if self.expected:
+            return "Could not ssh into sliver %s@%s"%(self.slicename,self.hostname)
+        else:
+            return "Could still ssh into sliver%s@%s (that was expected to be down)"%(self.slicename,self.hostname)
 
 class TestSlice:
 
@@ -24,111 +52,198 @@ class TestSlice:
             if(slice_spec['slice_fields']['name']== slice_name):
                 return slice_spec
 
-    def delete_slice(self):
+    def owner_auth(self):
         owner_spec = self.test_site.locate_user(self.slice_spec['owner'])
-        auth = TestUser(self,self.test_site,owner_spec).auth()
-        slice_fields = self.slice_spec['slice_fields']
-        slice_name = slice_fields['name']
-        self.test_plc.apiserver.DeleteSlice(auth,slice_fields['name'])
-        utils.header("Deleted slice %s"%slice_fields['name'])
+        return TestUser(self,self.test_site,owner_spec).auth()
 
-    
+    def slice_name (self):
+        return self.slice_spec['slice_fields']['name']
+
+    # init slice with people, and then add nodes 
     def create_slice(self):
-        owner_spec = self.test_site.locate_user(self.slice_spec['owner'])
-        auth = TestUser(self,self.test_site,owner_spec).auth()
+        auth = self.owner_auth()
         slice_fields = self.slice_spec['slice_fields']
         slice_name = slice_fields['name']
-
+        utils.header("Creating slice %s"%slice_name)
         self.test_plc.apiserver.AddSlice(auth,slice_fields)
         for username in self.slice_spec['usernames']:
                 user_spec=self.test_site.locate_user(username)
                 test_user=TestUser(self,self.test_site,user_spec)
                 self.test_plc.apiserver.AddPersonToSlice(auth, test_user.name(), slice_name)
+        # add initscript code or name as appropriate
+        if self.slice_spec.has_key('initscriptcode'):
+            iscode=self.slice_spec['initscriptcode']
+            utils.header("Adding initscript code %s in %s"%(iscode,slice_name))
+            self.test_plc.apiserver.AddSliceTag(self.test_plc.auth_root(), slice_name,'initscript_code',iscode)
+        elif self.slice_spec.has_key('initscriptname'):
+            isname=self.slice_spec['initscriptname']
+            utils.header("Adding initscript name %s in %s"%(isname,slice_name))
+            self.test_plc.apiserver.AddSliceTag(self.test_plc.auth_root(), slice_name,'initscript',isname)
+        if 'omf-friendly' in self.slice_spec:
+            utils.header("Making slice %s OMF-friendly"%slice_name)
+            self.test_plc.apiserver.AddSliceTag(self.test_plc.auth_root(), slice_name,'vref','omf')
+            self.test_plc.apiserver.AddSliceTag(self.test_plc.auth_root(), slice_name,'omf_control','yes')
+# setting vref directly like this was useful for multi-arch tests long ago - see wifilab
+# however this should rather use other tags by now, so we drop this for now
+#        if self.slice_spec.has_key ('vref'):
+#            vref_value=self.slice_spec['vref']
+#            self.test_plc.apiserver.AddSliceTag(self.test_plc.auth_root(), slice_name,'vref',vref_value)
+        # epilogue
+        self.add_nodes()
+
+    def check_vsys_defaults (self, options, *args, **kwds):
+        "check vsys tags match PLC_VSYS_DEFAULTS"
+        auth = self.owner_auth()
+        slice_fields = self.slice_spec['slice_fields']
+        slice_name = slice_fields['name']
+        vsys_tags = self.test_plc.apiserver.GetSliceTags (auth,{'tagname':'vsys','name':slice_name})
+        values=[ st['value'] for st in vsys_tags ]
+        expected=self.test_plc.plc_spec['expected_vsys_tags']
+        result = set(values) == set(expected)
+        if not result:
+            print 'Check vsys defaults with slice %s'%slice_name
+            print 'Expected %s'%expected
+            print 'Got %s'%values
+        return result
 
+    # just add the nodes and handle tags
+    def add_nodes (self):
+        auth = self.owner_auth()
+        slice_name = self.slice_name()
         hostnames=[]
         for nodename in self.slice_spec['nodenames']:
             node_spec=self.test_site.locate_node(nodename)
-            test_node=TestNode(self,self.test_site,node_spec)
+            test_node=TestNode(self.test_plc,self.test_site,node_spec)
             hostnames += [test_node.name()]
         utils.header("Adding %r in %s"%(hostnames,slice_name))
         self.test_plc.apiserver.AddSliceToNodes(auth, slice_name, hostnames)
-        if self.slice_spec.has_key('initscriptname'):
-            isname=self.slice_spec['initscriptname']
-            utils.header("Adding initscript %s in %s"%(isname,slice_name))
-            self.test_plc.apiserver.AddSliceTag(self.test_plc.auth_root(), slice_name,'initscript',isname)
-        if self.slice_spec.has_key ('vref'):
-            vref_value=self.slice_spec['vref']
-            self.test_plc.apiserver.AddSliceTag(self.test_plc.auth_root(), slice_name,'vref',vref_value)
         
-    def locate_key(self):
-        # locate the first avail. key
-        found=False
+    # trash the slice altogether
+    def delete_slice(self):
+        auth = self.owner_auth()
+        slice_name = self.slice_name()
+        utils.header("Deleting slice %s"%slice_name)
+        self.test_plc.apiserver.DeleteSlice(auth,slice_name)
+
+    # keep the slice alive and just delete nodes
+    def delete_nodes (self):
+        auth = self.owner_auth()
+        slice_name = self.slice_name()
+        print 'retrieving slice %s'%slice_name
+        slice=self.test_plc.apiserver.GetSlices(auth,slice_name)[0]
+        node_ids=slice['node_ids']
+        utils.header ("Deleting %d nodes from slice %s"%\
+                          (len(node_ids),slice_name))
+        self.test_plc.apiserver.DeleteSliceFromNodes (auth,slice_name, node_ids)
+
+    def locate_private_key(self):
+        key_names=[]
         for username in self.slice_spec['usernames']:
             user_spec=self.test_site.locate_user(username)
-            for keyname in user_spec['keynames']:
-                key_spec=self.test_plc.locate_key(keyname)
-                test_key=TestKey(self.test_plc,key_spec)
-                publickey=test_key.publicpath()
-                privatekey=test_key.privatepath()
-                keyname=test_key.name()
-                if os.path.isfile(publickey) and os.path.isfile(privatekey):
-                    found=True
-        return (found,privatekey)
-
-    def check_slice(self,options,timeout_minutes=20,silent_minutes=4,period=15):
-        timeout = datetime.datetime.now()+datetime.timedelta(minutes=timeout_minutes)
-        graceout = datetime.datetime.now()+datetime.timedelta(minutes=silent_minutes)
+            key_names += user_spec['key_names']
+        return self.test_plc.locate_private_key_from_key_names (key_names)
+
+    # for TestPlc.slice_mapper__tasks
+    # i.e. returns a list of CompleterTasks that are merged into the same Completer run
+    # to avoid waiting for as many slices as the Plc has
+    # also the __doc__ lines are used for the TestPlc methods, e.g. just 'ssh_slice'
+    def ssh_slice__tasks (self, options, *args, **kwds):
+        "tries to ssh-enter the slice with the user key, to check for slice creation"
+        return self.ssh_tasks(options, expected=True, *args, **kwds)
+
+    # when we expect the slice is not reachable
+    def ssh_slice_off__tasks (self, options, *args, **kwds):
+        "tries to ssh-enter the slice with the user key, expecting it to be unreachable"
+        return self.ssh_tasks(options, expected=False, *args, **kwds)
+
+    def ssh_tasks(self,options, expected=True, command=None):
+#                     timeout_minutes=20,silent_minutes=10,period_seconds=15):
+#        timeout  = timedelta(minutes=timeout_minutes)
+#        graceout = timedelta(minutes=silent_minutes)
+#        period   = timedelta(seconds=period_seconds)
+        if not command:
+            command="echo hostname ; hostname; echo id; id; echo uname -a ; uname -a"
         # locate a key
-        (found,remote_privatekey)=self.locate_key()
-        if not found :
+        private_key=self.locate_private_key()
+        if not private_key :
             utils.header("WARNING: Cannot find a valid key for slice %s"%self.name())
             return False
 
         # convert nodenames to real hostnames
-        slice_spec = self.slice_spec
-        restarted=[]
-        tocheck=[]
-        for nodename in slice_spec['nodenames']:
+        if expected:    msg="ssh slice access enabled"
+        else:           msg="ssh slice access disabled"
+        utils.header("checking for %s -- slice %s"%(msg,self.name()))
+
+        tasks=[]
+        slicename=self.name()
+        dry_run = getattr(options,'dry_run',False)
+        for nodename in self.slice_spec['nodenames']:
             (site_spec,node_spec) = self.test_plc.locate_node(nodename)
-            tocheck.append(node_spec['node_fields']['hostname'])
-
-        utils.header("checking ssh access into slice %s on nodes %r"%(self.name(),tocheck))
-        utils.header("max timeout is %d minutes, silent for %d minutes (period is %s)"%\
-                         (timeout_minutes,silent_minutes,period))
-        while tocheck:
-            for hostname in tocheck:
-                (site_spec,node_spec) = self.test_plc.locate_hostname(hostname)
-                date_test_ssh = TestSsh (hostname,key=remote_privatekey,username=self.name())
-                if datetime.datetime.now() >= graceout:
-                    utils.header('Trying to enter into slice %s@%s'%(self.name(),hostname))
-                # this can be ran locally as we have the key
-                date = date_test_ssh.run("echo hostname ; hostname; echo id; id; echo uname -a ; uname -a")==0
-                if date:
-                    utils.header("Successfuly entered slice %s on %s"%(self.name(),hostname))
-                    tocheck.remove(hostname)
-                else:
-                    # real nodes will have been checked once in case they're up - skip if not
-                    if TestNode.is_real_model(node_spec['node_fields']['model']):
-                        utils.header("WARNING : Checking slice %s on real node %s skipped"%(self.name(),hostname))
-                        tocheck.remove(hostname)
-                    # nm restart after first failure, if requested 
-                    if options.forcenm and hostname not in restarted:
-                        utils.header ("forcenm option : restarting nm on %s"%hostname)
-                        restart_test_ssh=TestSsh(hostname,key="keys/key1.rsa")
-                        access=restart_test_ssh.actual_command('service nm restart')
-                        if (access==0):
-                            utils.header('nm restarted on %s'%hostname)
-                        else:
-                            utils.header('Failed to restart nm on %s'%(hostname))
-                        restarted.append(hostname)
-            if not tocheck:
-                # we're done
-                return True
-            if datetime.datetime.now() > timeout:
-                for hostname in tocheck:
-                    utils.header("FAILURE to ssh into %s@%s"%(self.name(),hostname))
-                return False
-            # wait for the period
-            time.sleep (period)
-        # for an empty slice
-        return True
+            tasks.append( CompleterTaskSliceSsh(self.test_plc,node_spec['node_fields']['hostname'],
+                                                slicename,private_key,command,expected,dry_run))
+        return tasks
+#        return Completer (tasks).run (timeout, graceout, period)
+
+    def ssh_slice_basics (self, options, *args, **kwds):
+        "the slice is expected to be UP and we just check a few simple sanity commands, including 'ps' to check for /proc"
+        overall=True
+        if not self.do_ssh_slice_once(options,expected=True,  command='true'): overall=False
+        if not self.do_ssh_slice_once(options,expected=False, command='false'): overall=False
+        if not self.do_ssh_slice_once(options,expected=False, command='someimprobablecommandname'): overall=False
+        if not self.do_ssh_slice_once(options,expected=True,  command='ps'): overall=False
+        if not self.do_ssh_slice_once(options,expected=False, command='ls /vservers'): overall=False
+        return overall
+
+    # pick just one nodename and runs the ssh command once
+    def do_ssh_slice_once(self,options,command,expected):
+        # locate a key
+        private_key=self.locate_private_key()
+        if not private_key :
+            utils.header("WARNING: Cannot find a valid key for slice %s"%self.name())
+            return False
+
+        # convert nodenames to real hostnames
+        slice_spec = self.slice_spec
+        nodename=slice_spec['nodenames'][0]
+        (site_spec,node_spec) = self.test_plc.locate_node(nodename)
+        hostname=node_spec['node_fields']['hostname']
+
+        if expected:    msg="%s to return TRUE from ssh"%command
+        else:           msg="%s to return FALSE from ssh"%command
+            
+        utils.header("checking %s -- slice %s on node %s"%(msg,self.name(),hostname))
+        (site_spec,node_spec) = self.test_plc.locate_hostname(hostname)
+        test_ssh = TestSsh (hostname,key=private_key,username=self.name())
+        full_command = test_ssh.actual_command(command)
+        retcod = utils.system (full_command,silent=True)
+        if getattr(options,'dry_run',None): return True
+        if expected:    success = retcod==0
+        else:           success = retcod!=0
+        if not success: utils.header ("WRONG RESULT for %s"%msg)
+        return success
+
+    # for TestPlc.slice_mapper__tasks
+    # check that /vservers/<> is present/deleted
+    def slice_fs_present__tasks (self, options): 
+        "checks that /vservers/<slicename> exists on the filesystem"
+        return self.check_rootfs_tasks(options,expected=True)
+    def slice_fs_deleted__tasks (self, options): 
+        "checks that /vservers/<slicename> has been properly wiped off"
+        return self.check_rootfs_tasks (options,expected=False)
+
+    def check_rootfs_tasks (self, options, expected):
+        # use constant admin key
+        local_key = "keys/key_admin.rsa"
+        node_infos = self.test_plc.all_node_infos()
+        rootfs="/vservers/%s"%self.name()
+        if expected:
+            failure_message = "Could not stat %s"%rootfs
+        else:
+            failure_message = "Sliver rootfs still present in %s"%rootfs
+        class CompleterTaskRootfs (CompleterTaskNodeSsh):
+            def __init__ (self, nodename, qemuname):
+                CompleterTaskNodeSsh.__init__(self,nodename, qemuname, local_key, expected=expected,
+                                              message=failure_message, command="ls -d %s"%rootfs)
+            def failure_epilogue (self):
+                utils.system(self.test_ssh.actual_command("ls -l %s; du -hs %s"%(rootfs,rootfs),dry_run=self.dry_run))
+        return [ CompleterTaskRootfs (nodename, qemuname) for (nodename,qemuname) in node_infos ]