java - 访问 NAT 后面的 FreePastry 程序

Question

我正在尝试连接到我的程序，该程序在 NAT 后面使用 FreePastry 但无法连接。 mIP 是我的公共(public) IP，mBootport 和 mBindport 是 50001。我已将路由器中的这些端口转发到我的计算机，但它仍然不起作用。我禁用了防火墙，但什么也没有。我断开了路由器并直接连接到互联网，但仍然无法正常工作。它唯一一次在我的本地网络上工作。因此，配置文件的代码中大部分都有问题，但我看不出有什么问题。

        Environment env = new Environment();

        InetSocketAddress bootaddress = new InetSocketAddress(mIP, mBootport);

        NodeIdFactory nidFactory = new RandomNodeIdFactory(env);

        PastryNodeFactory factory = new SocketPastryNodeFactory(nidFactory, mBindport, env);

        for (int curNode = 0; curNode < mNumNodes; curNode++) {
            PastryNode node = factory.newNode();

            NetworkHandler app = new NetworkHandler(node, mLog);
            apps.add(app);

            node.boot(bootaddress);

            synchronized(node) {
                while(!node.isReady() && !node.joinFailed()) {
                node.wait(500);

                    if (node.joinFailed()) {
                        throw new IOException("Could not join the FreePastry ring.  Reason:"+node.joinFailedReason()); 
                    }
                }
            }

            System.out.println("Finished creating new node: " + node);
            mLog.append("Finished creating new node: " + node + "\n");
        }

        Iterator<NetworkHandler> i = apps.iterator();
        NetworkHandler app = (NetworkHandler) i.next();
        app.subscribe();


public class NetworkHandler implements ScribeClient, Application {

    int seqNum = 0;

    CancellableTask publishTask;

    Scribe myScribe;

    Topic myTopic;

    JTextArea mLog;

    protected Endpoint endpoint;

    public NetworkHandler(Node node, JTextArea log) {
        this.endpoint = node.buildEndpoint(this, "myinstance");

        mLog = log;

        myScribe = new ScribeImpl(node,"myScribeInstance");

        myTopic = new Topic(new PastryIdFactory(node.getEnvironment()), "example topic");
        System.out.println("myTopic = "+myTopic);
        mLog.append("myTopic = "+myTopic + "\n");

        endpoint.register();
    }

    public void subscribe() {
        myScribe.subscribe(myTopic, this); 
    }
}

freepastry.params

# this file holds the default values for pastry and it's applications
# you do not need to modify the default.params file to override these values
# instead you can use your own params file to set values to override the 
# defaults.  You can specify this file by constructing your
# rice.environment.Environment() with the filename you wish to use
# typically, you will want to be able to pass this file name from the command 
# line

# max number of handles stored per routing table entry
pastry_rtMax = 1
pastry_rtBaseBitLength = 4

# leafset size
pastry_lSetSize = 24

# maintenance frequencies
pastry_leafSetMaintFreq = 60
pastry_routeSetMaintFreq = 900

# drop the message if pastry is not ready
pastry_messageDispatch_bufferIfNotReady = false

# number of messages to buffer while an app hasn't yet been registered
pastry_messageDispatch_bufferSize = 32

# FP 2.1 uses the new transport layer
transport_wire_datagram_receive_buffer_size = 131072
transport_wire_datagram_send_buffer_size = 65536
transport_epoch_max_num_addresses = 2
transport_sr_max_num_hops = 5

# proximity neighbor selection
transport_use_pns = true

# number of rows in the routing table to consider during PNS
# valid values are ALL, or a number
pns_num_rows_to_use = 10

# commonapi testing parameters

# direct or socket
commonapi_testing_exit_on_failure = true
commonapi_testing_protocol = direct
commonapi_testing_startPort = 5009
commonapi_testing_num_nodes = 10
# set this to specify the bootstrap node
#commonapi_testing_bootstrap = localhost:5009

# random number generator's seed, "CLOCK" uses the current clock time
random_seed = CLOCK

# sphere, euclidean or gt-itm
direct_simulator_topology = sphere
# -1 starts the simulation with the current time
direct_simulator_start_time = -1
#pastry_direct_use_own_random = true
#pastry_periodic_leafset_protocol_use_own_random = true
pastry_direct_gtitm_matrix_file=GNPINPUT
# the number of stubs in your network
pastry_direct_gtitm_max_overlay_size=1000
# the number of virtual nodes at each stub: this allows you to simulate multiple "LANs" and allows cheeper scaling
pastry_direct_gtitm_nodes_per_stub=1
# the factor to multiply your file by to reach millis.  Set this to 0.001 if your file is in microseconds.  Set this to 1000 if your file is in seconds.
pastry_direct_gtitm_delay_factor=1.0
#millis of the maximum network delay for the generated network topologies
pastry_direct_max_diameter=200
pastry_direct_min_delay=2
#setting this to false will use the old protocols which are about 200 times as fast, but may cause routing inconsistency in a real network.  Probably won't in a simulator because it will never be incorrect about liveness
pastry_direct_guarantee_consistency=true

# rice.pastry.socket parameters
# tells the factory you intend to use multiple nodes
# this causes the logger to prepend all entries with the nodeid
pastry_factory_multipleNodes = true
pastry_factory_selectorPerNode = false
pastry_factory_processorPerNode = false
# number of bootstap nodehandles to fetch in parallel
pastry_factory_bootsInParallel = 1

# the maximum size of a message
pastry_socket_reader_selector_deserialization_max_size = 1000000
# the maximum number of outgoing messages to queue when a socket is slower than the number of messages you are queuing
pastry_socket_writer_max_queue_length = 30
pastry_socket_writer_max_msg_size = 20480
pastry_socket_repeater_buffer_size = 65536
pastry_socket_pingmanager_smallPings=true
pastry_socket_pingmanager_datagram_receive_buffer_size = 131072
pastry_socket_pingmanager_datagram_send_buffer_size = 65536
# the time before it will retry a route that was already found dead
pastry_socket_srm_check_dead_throttle = 300000
pastry_socket_srm_proximity_timeout = 3600000
pastry_socket_srm_ping_throttle = 30000
pastry_socket_srm_default_rto = 3000
pastry_socket_srm_rto_ubound = 10000
pastry_socket_srm_rto_lbound = 50
pastry_socket_srm_gain_h = 0.25
pastry_socket_srm_gain_g = 0.125
pastry_socket_scm_max_open_sockets = 300
pastry_socket_scm_max_open_source_routes = 30
# the maximum number of source routes to attempt, setting this to 0 will 
# effectively eliminate source route attempts
# setting higher than the leafset does no good, it will be bounded by the leafset
# a larger number tries more source routes, which could give you a more accurate 
# determination, however, is more likely to lead to congestion collapse
pastry_socket_srm_num_source_route_attempts = 8
pastry_socket_scm_socket_buffer_size = 32768
# this parameter is multiplied by the exponential backoff when doing a liveness check so the first will be 800, then 1600, then 3200 etc...
pastry_socket_scm_ping_delay = 800
# adds some fuzziness to the pings to help prevent congestion collapse, so this will make the ping be advanced or delayed by this factor
pastry_socket_scm_ping_jitter = 0.1
# how many pings until we call the node faulty
pastry_socket_scm_num_ping_tries = 5
pastry_socket_scm_write_wait_time = 30000
pastry_socket_scm_backoff_initial = 250
pastry_socket_scm_backoff_limit = 5
pastry_socket_pingmanager_testSourceRouting = false
pastry_socket_increment_port_after_construction = true
# if you want to allow connection to 127.0.0.1, set this to true
pastry_socket_allow_loopback = false
# these params will be used if the computer attempts to bind to the loopback address, they will open a socket to this address/port to identify which network adapter to bind to
pastry_socket_known_network_address = yahoo.com
pastry_socket_known_network_address_port = 80
pastry_socket_use_own_random = true
pastry_socket_random_seed = clock
# force the node to be a seed node
rice_socket_seed = false

# the parameter simulates some nodes being firewalled, base on rendezvous_test_num_firewalled
rendezvous_test_firewall = false
# probabilistic fraction of firewalled nodes
rendezvous_test_num_firewalled = 0.3
# don't firewall the first node, useful for testing
rendezvous_test_makes_bootstrap = false

# FP 2.1 uses the new transport layer
transport_wire_datagram_receive_buffer_size = 131072
transport_wire_datagram_send_buffer_size = 65536

# NAT/UPnP settings
nat_network_prefixes = 127.0.0.1;10.;192.168.
# Enable and set this if you have already set up port forwarding and know the external address
#external_address = 123.45.67.89:1234
#enable this if you set up port forwarding (on the same port), but you don't 
#know the external address and you don't have UPnP enabled
#this is useful for a firwall w/o UPnP support, and your IP address isn't static
probe_for_external_address = true
# values how to probe
pastry_proxy_connectivity_timeout = 15000 
pastry_proxy_connectivity_tries = 3
# possible values: always, never, prefix (prefix is if the localAddress matches any of the nat_network_prefixes
# whether to search for a nat using UPnP (default: prefix)
nat_search_policy = prefix
# whether to verify connectivity (default: boot)
firewall_test_policy = never
# policy for setting port forwarding the state of the firewall if there is already a conflicting rule: overwrite, fail (throw exception), change (use different port) 
# you may want to set this to overwrite or fail on the bootstrap nodes, but most freepastry applications can run on any available port, so the default is change
nat_state_policy = change
# the name of the application in the firewall, set this if you want your application to have a more specific name
nat_app_name = freepastry
# how long to wait for responses from the firewall, in millis
nat_discovery_timeout = 5000
# how many searches to try to find a free firewall port
nat_find_port_max_tries = 10
# uncomment this to use UPnP NAT port forwarding, you need to include in the classpath: commons-jxpath-1.1.jar:commons-logging.jar:sbbi-upnplib-xxx.jar
nat_handler_class = rice.pastry.socket.nat.sbbi.SBBINatHandler
# hairpinning: 
# default "prefix" requires more bandwidth if you are behind a NAT.  It enables multiple IP 
# addresses in the NodeHandle if you are behind a NAT.  These are usually the internet routable address, 
# and the LAN address (usually 192.168.x.x)
# you can set this to never if any of the following conditions hold:
#  a) you are the only FreePastry node behind this address
#  b) you firewall supports hairpinning see
# http://scm.sipfoundry.org/rep/ietf-drafts/behave/draft-ietf-behave-nat-udp-03.html#rfc.section.6
nat_nodehandle_multiaddress = prefix

# if we are not scheduled for time on cpu in this time, we setReady(false)
# otherwise there could be message inconsistency, because
# neighbors may believe us to be dead.  Note that it is critical
# to consider the amount of time it takes the transport layer to find a 
# node faulty before setting this parameter, this parameter should be
# less than the minimum time required to find a node faulty
pastry_protocol_consistentJoin_max_time_to_be_scheduled = 15000

# in case messages are dropped or something, how often it will retry to 
# send the consistent join message, to get verification from the entire
# leafset
pastry_protocol_consistentJoin_retry_interval = 30000
# parameter to control how long dead nodes are retained in the "failed set" in
# CJP (see ConsistentJoinProtocol ctor) (15 minutes)
pastry_protocol_consistentJoin_failedRetentionTime = 900000
# how often to cleanup the failed set (5 mins) (see ConsistentJoinProtocol ctor)
pastry_protocol_consistentJoin_cleanup_interval = 300000
# the maximum number of entries to send in the failed set, only sends the most 
recent detected failures (see ConsistentJoinProtocol ctor)
pastry_protocol_consistentJoin_maxFailedToSend = 20

# how often we send/expect to be sent updates
pastry_protocol_periodicLeafSet_ping_neighbor_period = 20000
pastry_protocol_periodicLeafSet_lease_period = 30000

# what the grace period is to receive a periodic update, before checking 
# liveness
pastry_protocol_periodicLeafSet_request_lease_throttle = 10000

# how many entries are kept in the partition handler's table
partition_handler_max_history_size=20
# how long entries in the partition handler's table are kept
# 90 minutes
partition_handler_max_history_age=5400000
# what fraction of the time a bootstrap host is checked
partition_handler_bootstrap_check_rate=0.05
# how often to run the partition handler
# 5 minutes
partition_handler_check_interval=300000

# the version number of the RouteMessage to transmit (it can receive anything that it knows how to)
# this is useful if you need to migrate an older ring
# you can change this value in realtime, so, you can start at 0 and issue a command to update it to 1
pastry_protocol_router_routeMsgVersion = 1

# should usually be equal to the pastry_rtBaseBitLength
p2p_splitStream_stripeBaseBitLength = 4
p2p_splitStream_policy_default_maximum_children = 24
p2p_splitStream_stripe_max_failed_subscription = 5
p2p_splitStream_stripe_max_failed_subscription_retry_delay = 1000

#multiring
p2p_multiring_base = 2

#past
p2p_past_messageTimeout = 30000
p2p_past_successfulInsertThreshold = 0.5

#replication

# fetch delay is the delay between fetching successive keys
p2p_replication_manager_fetch_delay = 500
# the timeout delay is how long we take before we time out fetching a key
p2p_replication_manager_timeout_delay = 20000
# this is the number of keys to delete when we detect a change in the replica set
p2p_replication_manager_num_delete_at_once = 100
# this is how often replication will wake up and do maintainence; 10 mins
p2p_replication_maintenance_interval = 600000 
# the maximum number of keys replication will try to exchange in a maintainence message
p2p_replication_max_keys_in_message = 1000

#scribe
p2p_scribe_maintenance_interval = 180000
#time for a subscribe fail to be thrown (in millis)
p2p_scribe_message_timeout = 15000

#util
p2p_util_encryptedOutputStream_buffer = 32678

#aggregation
p2p_aggregation_logStatistics = true
p2p_aggregation_flushDelayAfterJoin = 30000
#5 MINS
p2p_aggregation_flushStressInterval = 300000 
#5 MINS
p2p_aggregation_flushInterval = 300000 
#1024*1024
p2p_aggregation_maxAggregateSize = 1048576 
p2p_aggregation_maxObjectsInAggregate = 25
p2p_aggregation_maxAggregatesPerRun = 2
p2p_aggregation_addMissingAfterRefresh = true
p2p_aggregation_maxReaggregationPerRefresh = 100
p2p_aggregation_nominalReferenceCount = 2
p2p_aggregation_maxPointersPerAggregate = 100
#14 DAYS
p2p_aggregation_pointerArrayLifetime = 1209600000
#1 DAY 
p2p_aggregation_aggregateGracePeriod = 86400000
#15 MINS 
p2p_aggregation_aggrRefreshInterval = 900000 
p2p_aggregation_aggrRefreshDelayAfterJoin = 70000
#3 DAYS
p2p_aggregation_expirationRenewThreshold = 259200000 
p2p_aggregation_monitorEnabled = false
#15 MINS
p2p_aggregation_monitorRefreshInterval = 900000 
#5 MINS
p2p_aggregation_consolidationDelayAfterJoin = 300000 
#15 MINS
p2p_aggregation_consolidationInterval = 900000 
#14 DAYS
p2p_aggregation_consolidationThreshold = 1209600000 
p2p_aggregation_consolidationMinObjectsInAggregate = 20
p2p_aggregation_consolidationMinComponentsAlive = 0.8
p2p_aggregation_reconstructionMaxConcurrentLookups = 10
p2p_aggregation_aggregateLogEnabled = true
#1 HOUR
p2p_aggregation_statsGranularity = 3600000
#3 WEEKS 
p2p_aggregation_statsRange = 1814400000 
p2p_aggregation_statsInterval = 60000
p2p_aggregation_jitterRange = 0.1

# glacier
p2p_glacier_logStatistics = true
p2p_glacier_faultInjectionEnabled = false
p2p_glacier_insertTimeout = 30000
p2p_glacier_minFragmentsAfterInsert = 3.0
p2p_glacier_refreshTimeout = 30000
p2p_glacier_expireNeighborsDelayAfterJoin = 30000
#5 MINS
p2p_glacier_expireNeighborsInterval = 300000 
#5 DAYS
p2p_glacier_neighborTimeout = 432000000 
p2p_glacier_syncDelayAfterJoin = 30000
#5 MINS
p2p_glacier_syncMinRemainingLifetime = 300000 
#insertTimeout
p2p_glacier_syncMinQuietTime = 30000 
p2p_glacier_syncBloomFilterNumHashes = 3
p2p_glacier_syncBloomFilterBitsPerKey = 4
p2p_glacier_syncPartnersPerTrial = 1
#1 HOUR
p2p_glacier_syncInterval = 3600000 
#3 MINUTES
p2p_glacier_syncRetryInterval = 180000 
p2p_glacier_syncMaxFragments = 100
p2p_glacier_fragmentRequestMaxAttempts = 0
p2p_glacier_fragmentRequestTimeoutDefault = 10000
p2p_glacier_fragmentRequestTimeoutMin = 10000
p2p_glacier_fragmentRequestTimeoutMax = 60000
p2p_glacier_fragmentRequestTimeoutDecrement = 1000
p2p_glacier_manifestRequestTimeout = 10000
p2p_glacier_manifestRequestInitialBurst = 3
p2p_glacier_manifestRequestRetryBurst = 5
p2p_glacier_manifestAggregationFactor = 5
#3 MINUTES
p2p_glacier_overallRestoreTimeout = 180000 
p2p_glacier_handoffDelayAfterJoin = 45000
#4 MINUTES
p2p_glacier_handoffInterval = 240000 
p2p_glacier_handoffMaxFragments = 10
#10 MINUTES
p2p_glacier_garbageCollectionInterval = 600000 
p2p_glacier_garbageCollectionMaxFragmentsPerRun = 100
#10 MINUTES
p2p_glacier_localScanInterval = 600000 
p2p_glacier_localScanMaxFragmentsPerRun = 20
p2p_glacier_restoreMaxRequestFactor = 4.0
p2p_glacier_restoreMaxBoosts = 2
p2p_glacier_rateLimitedCheckInterval = 30000
p2p_glacier_rateLimitedRequestsPerSecond = 3
p2p_glacier_enableBulkRefresh = true
p2p_glacier_bulkRefreshProbeInterval = 3000
p2p_glacier_bulkRefreshMaxProbeFactor = 3.0
p2p_glacier_bulkRefreshManifestInterval = 30000
p2p_glacier_bulkRefreshManifestAggregationFactor = 20
p2p_glacier_bulkRefreshPatchAggregationFactor = 50
#3 MINUTES
p2p_glacier_bulkRefreshPatchInterval = 180000 
p2p_glacier_bulkRefreshPatchRetries = 2
p2p_glacier_bucketTokensPerSecond = 100000
p2p_glacier_bucketMaxBurstSize = 200000
p2p_glacier_jitterRange = 0.1
#1 MINUTE
p2p_glacier_statisticsReportInterval = 60000 
p2p_glacier_maxActiveRestores = 3

#transport layer testing params
org.mpisws.p2p.testing.transportlayer.replay.Recorder_printlog = true

# logging
#default log level
loglevel = WARNING

#example of enabling logging on the endpoint:
#rice.p2p.scribe@ScribeRegrTest-endpoint_loglevel = INFO

logging_packageOnly = true

logging_date_format = yyyyMMdd.HHmmss.SSS
logging_enable=true

# 24 hours
log_rotate_interval = 86400000
# the name of the active log file, and the filename prefix of rotated log
log_rotate_filename = freepastry.log
# the format of the date for the rotating log
log_rotating_date_format = yyyyMMdd.HHmmss.SSS

# true will tell the environment to ues the FileLogManager
environment_logToFile = false
# the prefix for the log files (otherwise will be named after the nodeId)
fileLogManager_filePrefix = 
# the suffix for the log files
fileLogManager_fileSuffix = .log
# wether to keep the line prefix (declaring the node id) for each line of the log
fileLogManager_keepLinePrefix = false
fileLogManager_multipleFiles = true
fileLogManager_defaultFileName = main

# false = append true = overwrite
fileLogManager_overwrite_existing_log_file = false

# the amount of time the LookupService tutorial app will wait before timing out
# in milliseconds, default is 30 seconds
lookup_service.timeout = 30000
# how long to wait before the first retry
lookup_service.firstTimeout = 500

编辑:通过 wireshark 确认消息确实到达计算机 freepastry 只是不接受连接。

Answer 1

不确定您所说的“不工作”是什么意思。要测试客户端和服务器(位于 NAT 后面)之间的连接性，您只需在客户端执行类似“telnet mIP mBindport”的操作，假设您有一个 telnet 实用程序(Linux 和 Mac 上默认，您可以安装一个，就像你 Windows 上的 nc(“netcat”)。

如果端口转发设置正确，当与服务器建立 TCP 连接时，您应该会看到如下内容。

Connected to localhost.
Escape character is '^]'.

一旦 TCP session 设置正确，您可以停止“telnet”程序并使用您的真实客户端(在 java 中)与您的服务器对话，它应该可以正常工作。

如果未建立 TCP session ，您可能需要在服务器端进行检查。使用 wireshark 或 tcpdump 抓包过滤器“tcp 端口 50001”，并运行上面的 telnet 命令检查是否有 TCP 数据包进来。

如果 wireshark 或 tcpdump 中没有显示任何内容，则说明您的防火墙(如端口转发)设置不正确。

如果 TCP 数据包确实出现在 wireshark 或 tcpdump 中，那么您的服务器程序可能有问题。使用命令 (linux) 检查它绑定(bind)的 IP 地址:

netstat -antp | grep 50001

(在 Windows 上，命令略有不同)。

通常它应该绑定(bind)到 IP 地址 0.0.0.0(所有 ip)，如果没有，您应该检查它绑定(bind)到的 IP 是否具有到外界(NAT 之外)的连接/路由。

祝你好运。