nimbus_server_script.py

#!/usr/bin/python

"""*
 * Copyright 2009 University of Victoria
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 * http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND,
 * either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 *
 * AUTHOR - Adam Bishop - ahbishop@uvic.ca
 * 
 * For comments or questions please contact the above e-mail address 
 * or Ian Gable - igable@uvic.ca
 *
 * """


__VERSION__ = '1.0'
# This absolute path to the config file is REQUIRED
CONF_FILE = "/usr/local/nagios/libexec/monitoring_config.cfg"

import sys
import commands
import os
import logging
from cStringIO import StringIO
from optparse import OptionParser
import subprocess
import commands
import re
import socket
import time
import ConfigParser
import logging

# NAGIOS Plug-In API return code values

NAGIOS_RET_OK = 0
NAGIOS_RET_WARNING = 1
NAGIOS_RET_CRITICAL = 2
NAGIOS_RET_UNKNOWN = 3

# These locations need to be absolute so that when the 'ij' utility is opened in a subprocess
# it can still locate them

SQL_IP_SCRIPT = "libexec/nimbus_derby_used_ips.sql"
SQL_RUNNING_VM_SCRIPT = "libexec/nimbus_derby_running_vms.sql"

NIMBUS_CONF = "etc/nimbus/workspace-service/"
NIMBUS_NET_CONF = NIMBUS_CONF+"network-pools"
NIMBUS_PHYS_CONF = NIMBUS_CONF+"vmm-pools"

NIMBUS_PBS_CONF = NIMBUS_CONF+"pilot.conf"
NIMBUS_PBS_SUPPORT = NIMBUS_CONF+"other/resource-locator-ACTIVE.xml"

CONF_FILE_SECTION = "Nimbus_Monitoring"
NIMBUS_ADDRESS = "Nimbus_Server_Address"
NIMBUS_LOCATION = "Nimbus_Install_Location"
GLOBUS_LOCATION = "Globus_Install_Location"
SERVER_TMP_LOCATION = "Server_Tmp_Location"
NAGIOS_LOCATION = "Nagios_Location"
JAVA_LOCATION = "Java_Location"
IJ_LOCATION = "IJ_Location"
DERBY_LOCATION = "Derby_Location"

ConfigMapping = {}

class Logger:
    """ A class to encapsulate useful logging features and setup

    """
    def __init__(self, name, errorLogFile=None):

        self.logger = logging.getLogger(name)

        self.logger.setLevel(logging.INFO)
        formatter = logging.Formatter('%(asctime)s : %(name)s : %(levelname)s : %(message)s')

        nagiosObservableHndlr = logging.StreamHandler(sys.stdout)
        nagiosObservableHndlr.setLevel(logging.ERROR)
        nagiosObservableHndlr.setFormatter(formatter)
        if (errorLogFile != None):
            pass
         
        # Uncommenting this code causes these plug-ins to fail to report in Nagios... 

#        fileOutputHndlr = logging.FileHandler(errorLogFile)
 #       fileOutputHndlr.setFormatter(formatter)
  #      fileOutputHndlr.setLevel(logging.ERROR)

   #     self.logger.addHandler(fileOutputHndlr)
        self.logger.addHandler(nagiosObservableHndlr)

    def warning(self, msg):
        self.logger.warning(msg)

    def info(self, msg):
        self.logger.info(msg)

    def error(self, msg):
        self.logger.error(msg)

    def debug(self, msg):
        self.logger.debug(msg)


# This global method loads all the user configured options from the configuration file and saves them
# into the ConfigMapping dictionary
def loadNimbusConfig(logger):

    cfgFile = ConfigParser.ConfigParser()
    if(os.path.exists(CONF_FILE)):
        cfgFile.read(CONF_FILE)
        try:
            ConfigMapping[NIMBUS_ADDRESS] = cfgFile.get(CONF_FILE_SECTION,NIMBUS_ADDRESS,0)
            ConfigMapping[NIMBUS_LOCATION] = cfgFile.get(CONF_FILE_SECTION,NIMBUS_LOCATION,0)
            ConfigMapping[SERVER_TMP_LOCATION] = cfgFile.get(CONF_FILE_SECTION, SERVER_TMP_LOCATION,0)
            ConfigMapping[NAGIOS_LOCATION] = cfgFile.get(CONF_FILE_SECTION, NAGIOS_LOCATION,0)
            ConfigMapping[JAVA_LOCATION] = cfgFile.get(CONF_FILE_SECTION, JAVA_LOCATION,0)
            ConfigMapping[IJ_LOCATION] = cfgFile.get(CONF_FILE_SECTION,IJ_LOCATION,0)
            ConfigMapping[GLOBUS_LOCATION] = cfgFile.get(CONF_FILE_SECTION,GLOBUS_LOCATION,0)
            ConfigMapping[DERBY_LOCATION] = cfgFile.get(CONF_FILE_SECTION,DERBY_LOCATION,0)
        except ConfigParser.NoSectionError:
            logger.error ("Unable to locate "+CONF_FILE_SECTION+" section in conf file - Malformed config file?")
            sys.exit(NAGIOS_RET_CRITICAL)
        except ConfigParser.NoOptionError, nopt:
            logger.error( nopt.message+" of configuration file")
            sys.exit(NAGIOS_RET_CRITICAL)
    else:
        logger.error("Configuration file not found in Nagios Plug-ins directory! (/usr/local/nagios/libexec)")
        sys.exit(NAGIOS_RET_CRITICAL)

# This is a recursive function to write out an XML doc from the pluginOutput data structure.
# This method will create XML from an arbitrary data structure layout 
def _createXMLWorker(data, currentOutput):

    if (type(data) == type(dict())):
        for key in data.keys():
            currentOutput.write("<"+str(key)+">")
            _createXMLWorker(data[key], currentOutput)
            currentOutput.write("</"+str(key)+">")

    elif (type(data) == type(list())):
        for key in range(len(data)):
            _createXMLWorker(data[key], currentOutput)

    else:
        currentOutput.write(str(data))
        return

def pluginExitN(messageIdentifier, pluginInfo, returnCode):

    # This method should be the only exit point for all the plug-ins. This ensures that 
    # Nagios requirements are meant and performance data is properly formatted to work
    # with the rest of the code. 

    outputString = StringIO()
    outputString.write("<?xml version=\"1.0\" encoding=\"UTF-8\"?>")

    localIP = (socket.gethostbyaddr( socket.gethostname() ))[2][0]
    outputString.write("<HeadNode>")
    outputString.write("<PhysicalIP>"+localIP+"</PhysicalIP>")

    outputString.write("<"+messageIdentifier+" id=\""+messageIdentifier+localIP+"\">")
    _createXMLWorker(pluginInfo, outputString)
   
    outputString.write("</"+messageIdentifier+">")
    outputString.write("</HeadNode>")

    sys.stdout.write(messageIdentifier+" | "+ outputString.getvalue()+"\n")
    sys.exit(returnCode)

class PluginObject:    
    """The most 'senior' of the base classes. This class sets up appropriate logging mechanisms to 
    conform with Nagios' API and plug-in coding rules. 
    """

    def __init__(self, callingClass):
    
        self.logger = Logger(callingClass)
        loadNimbusConfig(self.logger)
        self.pluginOutput = {}

class PluginCmdLineOpts:
    """ This class acts as the "central dispatcher" for determining what resource will be reported back
    to Nagios. Command line parameters act as the switches and determine which of the above classes
    gets instantianted.
    """

    def __init__(self):
            # Parse command-line options.
            parser = OptionParser()
            parser.add_option("--HNConsistent", action="callback",help="Verify internal Derby database consistency", callback=HeadNodeDBConsistent())
            parser.add_option("--HNVmmPool", action="callback",help="Publish Nimbus VMM pool information", callback=HeadNodeVMMPools())
            parser.add_option("--HNpbsmem", action="callback",help="Publish PBS/Torque available memory information", callback=HeadNodePBSMemory())
            parser.add_option("--HNpbssupport", action="callback",help="Publish support for PBS/Torque Pilot Jobs", callback=HeadNodePBSSupport())
            parser.add_option("--HNServiceInfo", action="callback", help="General Service Information about this Head Node's setup", callback=HeadNodeServiceInfo())

            self.parser = parser

    # This method is also responsible for "picking" what resource to monitor via the appropriate
    # command line switches (which I need to define). I don't want a single, monolithic script
    # running for ALL the resources, since this waters down NAGIOS's monitoring capabilities
    # (since that would make only a single resource to monitor)
    # Instead, this one script will be executed multiple time with different commandline options
    # to facilitate the monitoring of the different resources independant of one another

    def validate(self):
        (options, args) = self.parser.parse_args()

class HeadNodeServiceInfo(PluginObject):
    """
         This plug-in provides basic service information about the middle-ware running on the Head Node
    """
    def __init__(self):
        self.resourceName="Service"
        PluginObject.__init__(self,self.__class__.__name__)
        
    def __call__(self, option, opt_str, value, parser):

        self.pluginOutput["HostName"] = str(os.uname()[1])
        self.pluginOutput["Type"] = "Nimbus"
        self.pluginOutput["Port"] = str(8443)
        self.pluginOutput["Path"] = "/wsrf/services/WorkspaceFactoryService"
        self.pluginOutput["IP"] = str((socket.gethostbyaddr( socket.gethostname() ))[2][0])

        pluginExitN(self.resourceName, self.pluginOutput, NAGIOS_RET_OK)

class HeadNodePBSSupport(PluginObject):
    """ This class parses the Nimbus configuration file that dictates whether Torque/PBS or Nimbus along controls
    job submission & queueing. The prescense of a certain string within the file is enough to define which of
    the two options is active. 
    """
    def __init__(self):
        self.resourceName = "PBS-Support"
        PluginObject.__init__(self, self.__class__.__name__)

    def __call__(self, option, opt_str, value, parser):

        try:
            fileHandle = open(ConfigMapping[NIMBUS_LOCATION]+NIMBUS_PBS_SUPPORT)
        except IOError:
            self.logger.error("IOError opening resource-locator-ACTIVE.xml for processing!")
            sys.exit(NAGIOS_RET_CRITICAL)
        else:
            pbsSupported = False
            for line in fileHandle:
                if(line.find(".pilot.PilotSlotManagement") == -1):
                    continue
                else:
                    pbsSupported = True
            fileHandle.close()    
            self.pluginOutput["PilotJobsActive"] = str(pbsSupported)
            pluginExitN(self.resourceName,self.pluginOutput, NAGIOS_RET_OK)

class HeadNodePBSMemory(PluginObject):
    """ The class parses the Nimbus configuration files to determine how much total memory is available
    to Torque/PBS jobs (should Torque/PBS be active). This only has meaning of course if Torque/PBS
    is the active scheduling & queueing component. Note that this resource replaces the VMM-Pools
    memory counts (that is they are not used should Torque/PBS be the queuer & scheduler)
    """
    def __init__(self):
        self.resourceName = "PBS-Memory"
        PluginObject.__init__(self, self.__class__.__name__)

    def __call__(self, option, opt_str, value, parser):

        try:
            fileHandle = open(ConfigMapping[NIMBUS_LOCATION]+NIMBUS_PBS_CONF)
        except IOError:
            self.logger.error("IOError opening pilot.conf for processing!")
            sys.exit(NAGIOS_RET_CRITICAL)
        else:    
            memory = 0
            for line in fileHandle:

                if(line.find("memory.maxMB") == -1):
                    continue
                else:
                    lineSegs = line.split("=")
                    memory = int(lineSegs[1].strip())
            
            fileHandle.close()
            self.pluginOutput["TotalMemoryMB"] = str(memory)
            pluginExitN(self.resourceName, self.pluginOutput, NAGIOS_RET_OK)
            
            
class HeadNodeDBConsistent(PluginObject):
    """ This is a complicated class for attempting to solve the problem of determing if the 
    Nimbus DerbyDB embedded within itself is "consistent". "Consistent" isn't rigorously defined 
    anywhere, which creates a bit of a challenge.

    So, this plug-in utilizes the 'ij' utility included with the DerbyDB distribution. A static
    text file with the requisite connection string & query is redirected from STDIN to the ij
    utility to perform a query to find all the used IP addresses recorded. These addresses are
    then 'pinged' to see if they are alive. This of course relies on VMs having proper networking
    configurations & allow ICMP pings to reach them. Should these addresses not be reachable, it 
    is assumed the DB is in an inconsistent state and the plug-in reports an error.

    After the above "first pass" the 'ij' utility and another static connect/query file is utilized
    to find all the used IP addresses and the Startup, shutdown & termination times. The shutdown time
    is then compared to the current time to see if we have passed the time when shutdown should occur. 
    If we have passed the shutdown time, then the IP address is again 'pinged' to see if the VM is 
    still alive. If it is, then the DB is considered to be inconsistent and the plug-in report the
    error.

    This obviously isn't an ideal way of determining 'consistency', but its the best fit given the 
    VMs should all have proper networking setup to function properly
    """
    def __init__(self):
        self.resourceName = "IaasDiagnostics"
        PluginObject.__init__(self,self.__class__.__name__)
    
    def __call__(self, option, opt_str, value, parser):
    
        isConsistent = True

        query = ConfigMapping[IJ_LOCATION]+ " "+ConfigMapping[NAGIOS_LOCATION]+SQL_IP_SCRIPT
        #print query
        output,status = (subprocess.Popen([query],stdout = subprocess.PIPE, stderr = subprocess.PIPE, shell=True, env={'DERBY_HOME':ConfigMapping[DERBY_LOCATION],'JAVA_HOME':ConfigMapping[JAVA_LOCATION],'GLOBUS_HOME':ConfigMapping[NIMBUS_LOCATION]})).communicate()
	
        if(status != ""):
            self.logger.error("An error occured querying DerbyDB with IJ "+ status)
        
        derbyIPs = []
        patt = re.compile(r"(\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3})") 
        for line in output.split():
            myRe = patt.search(line)
            if(myRe):
                derbyIPs.append({line.strip(): False})    
        # So now I know what IPs should be reachable    
        for remoteVM in derbyIPs:
            # There should be only 1 key in the dictionary structure
            # the '[0]' extracts the IP address as a string
            addr = remoteVM.keys()[0]
            if(self.ping(addr)):
                derbyIPs[derbyIPs.index(remoteVM)][addr] = True    

        for foundVM in derbyIPs:
            # Again, there should be only 1 'value' for a given key
            # and the 'value' is a True/False boolean
            if not (foundVM.values()[0]):
                self.logger.error("Unable to reach VM via PING - "+foundVM.keys()[0])
                isConsistent = False
        # OK, so now are their reachable VMs that should be terminated?

        query = ConfigMapping[IJ_LOCATION] + " "+ConfigMapping[NAGIOS_LOCATION] + SQL_RUNNING_VM_SCRIPT
        output,status = (subprocess.Popen([query],stdout = subprocess.PIPE, stderr = subprocess.PIPE, shell=True, env={'DERBY_HOME':ConfigMapping[DERBY_LOCATION],'JAVA_HOME':ConfigMapping[JAVA_LOCATION],'GLOBUS_HOME':ConfigMapping[NIMBUS_LOCATION]})).communicate()

        if(status != ""):
            self.logger.error("An error occured querying DerbyDB with IJ "+ status)

        ijOutput = output.split()
        counter = 0
        VMs = []
        for line in ijOutput:

                myRe = patt.search(line)
                if(myRe):
                    network = line.split('|')[0]
                    networkEntries = network.split(';')
                    # The networking info and termination time share the same line in the returned results from the IJ query
                    # so an array is returned by the split, and the term time is in the second entry
                    # NOTE - All the times stored in the database have 3 additional #'s of precision compared to the Unix
                    # time command of 'date +%s', which gives us the time since epoch
                    # Thus, the trailing 3 numerals of all time values are stripped off

                    termTime = (line.split('|')[1])[:-3]
                    # Both 'StartTime' and 'ShutdownTime' have a pesky pipe '|' char preceding the number that needs to be
                    # stripped off
                    # Also, various fancy array slicing is used to cut the various timestamps into a standard form
                    startTime = ((ijOutput[counter+1])[1:len(ijOutput[counter+1])])[:-3]
                    shutdownTime = ((ijOutput[counter+2])[1:len(ijOutput[counter+2])])[:-3]
                
                    VMs.append({"IP":networkEntries[5], "TermTime":termTime,"StartTime":startTime, "ShutdownTime":shutdownTime})
                counter += 1
        currentTime = int(time.time())
        for entry in VMs:
            if(int(currentTime) > int(entry["ShutdownTime"] )):
                # Need to ping the VM to see if it's still alive
                if(self.ping(entry["IP"])  ):
                    self.logger.error("VM at IP address: "+entry["IP"]+" is alive and shouldn't be!")
                else:
                    self.logger.error("VM at IP address: "+entry["IP"]+" in DB but unreachable!")    
                isConsistent = False  
              
        if (isConsistent):
            retCode = NAGIOS_RET_OK
            idString = "Consistent"
        else:
            retCode = NAGIOS_RET_CRITICAL
            idString = "Inconsistent"
        self.pluginOutput["InternalRepresentation"] = idString
        pluginExitN(self.resourceName,self.pluginOutput, retCode)

    # The only "tunable" paramter for this function is the number value that comes after the '-c' in the Popen
    # command below. This is a parameter for the 'ping' command line utility and dictates the number of 'pings'
    # sent to the target. I have it set to '1' for speed reasons, as this function is somehwat slow in terms of
    # execution time. Adjust this value to send more pings if required        
    def ping(self, hostaddress):
       
        ping = subprocess.Popen(["ping","-c","1",hostaddress],stdout = subprocess.PIPE,stderr = subprocess.PIPE)
        out, error = ping.communicate()
        
        retVal = True
        if(out.find("Unreachable") != -1):
            retVal = False
        return retVal


class HeadNodeVMMPools(PluginObject):
    """ This class parses the Nimbus VMM Pool configuration files to build a representation of how much memory is 
        available to each defined VMM pool. This is done by parsing the appropriate configuration files in the 
        Nimbus distribution. Names of pools and memory allocated to each also requires accessing the network-conf
        data to provide intersections in the data.
    """      
    def __init__(self):
        PluginObject.__init__(self,self.__class__.__name__)
        self.resourceName = "VMM-Pools"

    def __call__(self, option, opt_str, value, parser):
       
        try:
            vmmPools = os.listdir(str(ConfigMapping[NIMBUS_LOCATION])+NIMBUS_PHYS_CONF)
        except OSError, ose:
            self.logger.error("Error listing VMMPool directory: "+str(ose))  
            sys.exit(NAGIOS_RET_CRITICAL)
        try:
            netPools = os.listdir(ConfigMapping[NIMBUS_LOCATION]+NIMBUS_NET_CONF)
        except OSError, ose:
           self.logger.error("Error listing NetPool directory: "+str(ose))
           sys.exit(NAGIOS_RET_CRITICAL)

        poolListing = {}
        for pool in vmmPools:
            
            # Ignore "dot" file/folders - hidden directories
            if(pool.startswith(".")):
                continue
            totalNetPools = {"ANY":0}
            # Initialize the appropriate pool names to 0 MB memory
            for npool in netPools:
                if(npool.startswith(".")):
                    continue
                totalNetPools.update({npool:0})
            try:
                fileHandle = open(ConfigMapping[NIMBUS_LOCATION]+NIMBUS_PHYS_CONF+"/"+pool)
                workerNodes = []
                for entry in fileHandle:
                    # Skip comments and whitespace
                    if(entry.startswith("#") or entry.isspace()):
                        continue
                    t = entry.split()
                    workerNodes.append(t)
                fileHandle.close()
                for entry in workerNodes:
                    # IF there is only 2 entries on this given line, that means
                    # the particular workerNode has no specific network pool 
                    # configured, so it's memory count gets added to the "global"
                    # or DEFAULT count
                    keyList = []
                    if(len(entry)< 3):
                        keyList.append("ANY")
                    else:

                        keyList = entry[2].split(",")

                    for network in keyList:
                        # * == "ANY" in the VMM Configuration file
                        if( network == "*"):
                            totalNetPools["ANY"] += int(entry[1])
                            continue
                        # Increment the memory count for this network                        
                        if network in (totalNetPools.keys()):
                            totalNetPools[network] += int(entry[1])            
                        else:
                            self.logger.error("Erroneous entry in the VMM configuration: "+ network+" - Ignoring")
                poolListing[pool] = totalNetPools
            except IOError:
                self.logger.error("Error opening VMM-pool: "+ConfigMapping[GLOBUS_LOCATION]+NIMBUS_PHYS_CONF+ pool)
                sys.exit(NAGIOS_RET_CRITICAL)
        self.pluginOutput= []
        for key in poolListing.keys():
            for entry in totalNetPools.keys():
                # This structure is needed to ensure the outout XML conforms to the public Schema
                tdict = {"Name":key+"-"+entry, "MemorySizeMB": str(poolListing[key][entry])}
                self.pluginOutput.append({"Pool":tdict}) 

        pluginExitN(self.resourceName, self.pluginOutput,  NAGIOS_RET_OK)


if __name__ == '__main__':

    testObject = PluginCmdLineOpts()
    testObject.validate()

    # This sys.exit call should NEVER be reached under normal circumstances, or any....
    sys.exit(NAGIOS_RET_CRITICAL)