mpdcheck.py

mpi并行计算的c++代码 可用vc或gcc编译通过 可以用来搭建并行计算试验环境

        print msg.replace('        ','')
    
    fipaddr1 = 0
    try:
        ghbnf = gethostbyname_ex(fqhn1)
        if verbose:
            print "gethostbyname_ex: ", ghbnf
        fipaddr1 = ghbnf[2][0]
        if fipaddr1.startswith('127'):
            msg = """
            **********
            Your fully qualified hostname resolves to 127.0.0.1, which
            is the IP address reserved for localhost. This likely means
            that you have a line similar to this one in your /etc/hosts file:
                 127.0.0.1   $fqhn
            This should be perhaps changed to the following:
                 127.0.0.1   localhost.localdomain localhost
            **********
            """
        try:
            ghbaf = gethostbyaddr(fipaddr1)
        except:
            print "*** gethostbyaddr failed for this hosts's IP %s" % (uipaddr1)
    except:
        if long_messages:
            msg = """
            **********
            The system call gethostbyname(3) failed to resolve your
            fully qualified hostname, or $fqhn. This can be caused by
            missing info from your /etc/hosts file or your system not
            having correctly configured name resolvers, or by your IP 
            address not existing in resolution services.
            If you run DNS, please check and make sure that your
            DNS server has the correct forward A record set up for yout
            machine's hostname. If you are not using DNS and are only using
            hosts files, please check that a line similar to the one below
            exists in your /etc/hosts file:
                $ipaddr $fqhn
            If you intend to use DNS but you are not sure that it is
            correctly configured, please check that the file /etc/resolv.conf
            contains entries similar to the following:
                nameserver 1.2.3.4
            where 1.2.3.4 is an actual IP of one of your nameservers.
            **********
            """
        else:
            msg = "*** gethostbyname_ex failed for host %s" % (fqhn1)
        print msg.replace('        ','')
    
    if verbose:
        print 'checking that IP addrs resolve to same host'
    if uipaddr1 and fipaddr1 and uipaddr1 != fipaddr1:
        msg = """
            **********
            Your fully qualified and unqualified names do not resolve to
            the same IP. This likely means that your DNS domain name is not
            set correctly.  This might be fixed by adding a line similar
            to the following to your /etc/hosts:
                 $ipaddr             $fqhn   $uqdn
            **********
            """
        print msg.replace('        ','')
    
    
    if verbose:
        print 'now do some gethostbyaddr and gethostbyname_ex for machines in hosts file'
    # See if we can do gethostXXX, etc. for hosts in hostsFromFile
    for host in hostsFromFile:
        uqhn2 = host
        fqhn2 = getfqdn(uqhn2)
        uipaddr2 = 0
        if verbose:
            print 'checking gethostbyXXX for unqualified %s' % (uqhn2)
        try:
            ghbnu = gethostbyname_ex(uqhn2)
            if verbose:
                print "gethostbyname_ex: ", ghbnu
            uipaddr2 = ghbnu[2][0]
            try:
                ghbau = gethostbyaddr(uipaddr2)
            except:
                print "*** gethostbyaddr failed for remote hosts's IP %s" % (fipaddr2)
        except:
            print "*** gethostbyname_ex failed for host %s" % (fqhn2)
        if verbose:
            print 'checking gethostbyXXX for qualified %s' % (uqhn2)
        try:
            ghbnf = gethostbyname_ex(fqhn2)
            if verbose:
                print "gethostbyname_ex: ", ghbnf
            fipaddr2 = ghbnf[2][0]
            if uipaddr2  and  fipaddr2 != uipaddr2:
                print "*** ipaddr via uqn (%s) does not match via fqn (%s)" % (uipaddr2,fipaddr2)
            try:
                ghbaf = gethostbyaddr(fipaddr2)
            except:
                print "*** gethostbyaddr failed for remote hosts's IP %s" % (fipaddr2)
        except:
            print "*** gethostbyname_ex failed for host %s" % (fqhn2)
    
    
    # see if we can run /bin/date on remote hosts
    if not do_ssh:
        exit(0)
    
    for host in hostsFromFile:
        cmd = "ssh %s -x -n /bin/echo hello" % (host)
        if verbose:
            print 'trying: %s' % (cmd)
        runner = Popen3(cmd,1,0)
        runout = runner.fromchild
        runerr = runner.childerr
        runin  = runner.tochild
        runpid = runner.pid
        try:
            (readyFDs,unused1,unused2) = select([runout],[],[],9)
        except Exception, data:
            print 'select 1 error: %s ; %s' % ( data.__class__, data)
            exit(-1)
        if len(readyFDs) == 0:
            print '** ssh timed out to %s' % (host)
        line = ''
        failed = 0
        if runout in readyFDs:
            line = runout.readline()
            if not line.startswith('hello'):
                failed = 1
        else:
            failed = 1
        if failed:
            print '** ssh failed to %s' % (host)
            print '** here is the output:'
            if line:
                print line,
            done = 0
            fds = [runout,runerr]
            while not done:
                try:
                    (readyFDs,unused1,unused2) = select(fds,[],[],1)
                except Exception, data:
                    print 'select 2 error: %s ; %s' % ( data.__class__, data)
                    exit(-1)
                if runout in readyFDs:
                    line = runout.readline()
                    if line:
                        print line,
                    else:
                        fds.remove(runout)
                elif runerr in readyFDs:
                    line = runerr.readline()
                    if line:
                        print line,
                    else:
                        fds.remove(runerr)
                else:
                    done = 1
        try:
            kill(runpid,SIGKILL)
            runout.close()
            runerr.close()
            runin.close()
        except:
            pass
        if failed:
            exit(-1)
    
    # see if we can run mpdcheck on remote hosts
    for host in hostsFromFile:
        cmd1 = path.join(fullDirName,'mpdcheck.py') + ' -s'
        if verbose:
            print 'starting server: %s' % (cmd1)
        runner1 = Popen3(cmd1,1,0)
        runout1 = runner1.fromchild
        runerr1 = runner1.childerr
        runin1  = runner1.tochild
        runpid1 = runner1.pid
        try:
            (readyFDs,unused1,unused2) = select([runout1],[],[],9)
        except Exception, data:
            print 'select 3 error: %s ; %s' % ( data.__class__, data)
            exit(-1)
        if len(readyFDs) == 0:
            print '** timed out waiting for local server to produce output'
        line = ''
        failed = 0
        port = 0
        if runout1 in readyFDs:
            line = runout1.readline()
            if line.startswith('server listening at '):
                port = line.rstrip().split(' ')[-1]
            else:
                failed = 1
        else:
            failed = 1
        if failed:
            print 'could not start mpdcheck server'
            print 'here is the output:'
            if line:
                print line,
            done = 0
            fds = [runout1,runerr1]
            while not done:
                try:
                    (readyFDs,unused1,unused2) = select(fds,[],[],1)
                except Exception, data:
                    print 'select 4 error: %s ; %s' % ( data.__class__, data)
                    exit(-1)
                if runout in readyFDs:
                    line = runout.readline()
                    if line:
                        print line,
                    else:
                        fds.remove(runout)
                elif runerr in readyFDs:
                    line = runerr.readline()
                    if line:
                        print line,
                    else:
                        fds.remove(runerr)
                else:
                    done = 1
        if failed:
            try:
                kill(runpid1,SIGKILL)
            except:
                pass
            exit(-1)
        cmd2 = "ssh %s -x -n %s%smpdcheck.py -c %s %s" % (host,fullDirName,path.sep,fqhn1,port)
        if verbose:
            print 'starting client: %s' % (cmd2)
        runner2 = Popen3(cmd2,1,0)
        runout2 = runner2.fromchild
        runerr2 = runner2.childerr
        runin2  = runner2.tochild
        runpid2 = runner2.pid
        try:
            (readyFDs,unused1,unused2) = select([runout2],[],[],9)
        except Exception, data:
            print 'select 3 error: %s ; %s' % ( data.__class__, data)
            exit(-1)
        if len(readyFDs) == 0:
            print '** timed out waiting for client on %s to produce output' % (host)
        line = ''
        failed = 0
        port = 0
        if runout2 in readyFDs:
            line = runout2.readline()
            if not line.startswith('client successfully recvd'):
                failed = 1
        else:
            failed = 1
        if failed:
            print 'client on %s failed to access the server' % (host)
            print 'here is the output:'
            if line:
                print line,
            done = 0
            fds = [runout2,runerr2]
            while not done:
                try:
                    (readyFDs,unused1,unused2) = select(fds,[],[],1)
                except Exception, data:
                    print 'select 4 error: %s ; %s' % ( data.__class__, data)
                    exit(-1)
                if runout2 in readyFDs:
                    line = runout2.readline()
                    if line:
                        print line,
                    else:
                        fds.remove(runout2)
                elif runerr2 in readyFDs:
                    line = runerr2.readline()
                    if line:
                        print line,
                    else:
                        fds.remove(runerr2)
                else:
                    done = 1
        try:
            kill(runpid2,SIGKILL)
        except:
            pass
        if failed:
            try:
                kill(runpid1,SIGKILL)
            except:
                pass
            exit(-1)