[SNMP4J] SNMP4J Correct Multi-thread Design for use it (under Linux) ?
TESSALONIKOS, THIERRY
thierry.tessalonikos at atos.net
Fri Sep 23 15:57:31 CEST 2016
Hi,
I need to use SNMP4J API in multi-threading design, because I must send 2 snmp set in parallel toward 2 Network Elements for sending them a parameters...
So, I have a main thread that initialize the SNMP4J API, with a singleton like this :
public class SNMP4JInitialize implements SNMPEngineInitInterface{
// properties SNMP4J
private static SNMP4JInitialize snmp4JInitializeInstance;
private static TransportMapping transport;
private static Snmp snmp;
private static USM usm;
private static OctetString localEngineID;
private SNMP4JInitialize() throws IOException {
// Intialisation SNMP4J
transport = new DefaultUdpTransportMapping(); //@TODO utiliser un autre constructeur pour positionner adresse IP d'écoute
snmp = new Snmp(transport);
localEngineID = new OctetString(MPv3.createLocalEngineID());
usm = new USM(SecurityProtocols.getInstance(), localEngineID, 0);
SecurityModels.getInstance().addSecurityModel(usm);
SecurityProtocols.getInstance().addPrivacyProtocol(new PrivAES256With3DESKeyExtension());
transport.listen();
// Instanaciation de la classe qui gere les UsmUer automatiquement à partir des MAJ JMS envoyées par l'appli JEE
OctetString securityName1 = new OctetString("toto");
UsmUser usmUser1 = new UsmUser(securityName1, // securityName
AuthMD5.ID, // authenticationProtocol
new OctetString("toto123456"), // authenticationPassphrase
PrivAES128.ID, // privacyProtocol
new OctetString("toto123456")); // privacyPassphrase
snmp.getUSM().addUser(usmUser1);
OctetString securityName2 = new OctetString("titi");
UsmUser usmUser2 = new UsmUser(securityName2, // securityName
AuthSHA.ID, // authenticationProtocol
new OctetString("toto123456"), // authenticationPassphrase
PrivAES256With3DESKeyExtension.ID, // privacyProtocol
//PrivAES256.ID, // privacyProtocol
new OctetString("toto123456")); // privacyPassphrase
snmp.getUSM().addUser(usmUser2);
}
public static Snmp getSnmp() {
return snmp;
}
public static SNMP4JInitialize getInstance() throws IOException, Exception {
if (snmp4JInitializeInstance == null) {
snmp4JInitializeInstance = new SNMP4JInitialize();
}
return (snmp4JInitializeInstance);
}
}
And after I launch 2 thread in parallel (from a thread pool, managed by ExecutorService class...)
Each thread create a new UserTarget and getSNMP instance from the Initialization Singleton, and send the PDU with snmp.send(pdu, target) like this :
public class SNMP4JRequestSender implements SNMPModuleSenderInterface {
public RequestTaskStatus sendsnmpv3Request(RequestTaskStatus requestTaskStatus, RequestTaskSNMP requestTaskSNMP) throws IOException {
final Address targetAddress = GenericAddress.parse(requestTaskSNMP.getSnmpTargetURL());
// create the target
UserTarget target = new UserTarget();
target.setAddress(targetAddress);
target.setRetries(requestTaskSNMP.getRetry());
target.setTimeout(requestTaskSNMP.getTimout());
target.setVersion(SnmpConstants.version3);
target.setSecurityLevel(SecurityLevel.AUTH_PRIV);
target.setSecurityName(requestTaskSNMP.getSecurityName()); // The same name used for usmUser
//create the PDU
ScopedPDU pdu = new ScopedPDU();
List<VariableBinding> variableBindings = requestTaskSNMP.getOids();
if ((variableBindings == null) || (variableBindings.isEmpty())) {
requestTaskStatus.setResultStatus(GeneralResultStatusEnum.Failed);
requestTaskStatus.setReasonStatus(ReasonStatusEnum.NoOIDinSNMPRequest);
return (requestTaskStatus);
}
for (VariableBinding variableBinding : variableBindings) {
pdu.add(variableBinding);
}
pdu.setType(requestTaskSNMP.getPDUType());
// send the PDU
final ResponseEvent response = SNMP4JInitialize.getSnmp().send(pdu, target);
// extract the response PDU (could be null if timed out)
final PDU responsePDU = response.getResponse();
if (responsePDU == null) {
requestTaskStatus.setResultStatus(GeneralResultStatusEnum.Failed);
requestTaskStatus.setReasonStatus(ReasonStatusEnum.NoResponseFromNetworkElement);
}
return (requestTaskStatus);
}
}
Is this design is correct or supported with SNMP4J API ? (thread safe, ...).
It seems work well when I launch in parallel the 2 thread for sending 2 snmp request toward to different IP address target.
But when I launch the 2 thread toward the same IP address (but each thread use its own instance of UserTarget, but with the same IP) a have the following error only for the second request :
- org.snmp4j.MessageException : Operation not permitted
In this context, I observe with wireshark that the targets responds with SNMP Report : usmStatsNotInTimeWindows.0 but at my class SNMP4JInitalize is a Singleton, I do not recreate new instances of snmp() class or I do not reinit the EngineID...
I have this problem only one time, at the first run, I a second run I do not have any error. Is usmStatsNotInTimeWindows.0 is normal the first time for initializing the global instance of EngineID ?
Wireshark trace with iptable permit all (iptables -F) :
No. Time Source Destination Protocol Length Info
1 2016-09-23 13:26:35.033849965 192.168.1.3 192.168.1.4 SNMP 103 get-request
2 2016-09-23 13:26:35.034764515 192.168.1.4 192.168.1.3 SNMP 141 report SNMP-USER-BASED-SM-MIB::usmStatsUnknownEngineIDs.0
3 2016-09-23 13:26:35.214557739 192.168.1.3 192.168.1.4 SNMP 180 get-request SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16975104
4 2016-09-23 13:26:35.215829657 192.168.1.4 192.168.1.3 SNMP 157 report SNMP-USER-BASED-SM-MIB::usmStatsNotInTimeWindows.0
5 2016-09-23 13:26:35.216503803 192.168.1.3 192.168.1.4 SNMP 181 get-request SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16975104
6 2016-09-23 13:26:35.218085613 192.168.1.4 192.168.1.3 SNMP 185 get-response SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16975104
7 2016-09-23 13:26:40.037023750 192.168.1.3 192.168.1.4 SNMP 181 get-request SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16974080
8 2016-09-23 13:26:40.043708644 192.168.1.4 192.168.1.3 SNMP 185 get-response SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16974080
SNMP4J do not receive : get-response SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16974080 and instead : org.snmp4j.MessageException : Operation not permitted.
I also observe that the response from the second target is OK, but seems to be blocked by iptable. But my iptable permit all. So, when I completely stop the firewalld, this works with no error, but in observe SNMP Report : usmStatsNotInTimeWindows.0 only for the 2 first request :
Wirewhark trace without firewalld : no error SNMP4J but Report : usmStatsNotInTimeWindows.0
No. Time Source Destination Protocol Length Info
1 2016-09-23 13:41:22.582650149 192.168.1.3 192.168.1.4 SNMP 103 get-request
2 2016-09-23 13:41:22.582658266 192.168.1.3 192.168.1.4 SNMP 103 get-request
3 2016-09-23 13:41:22.583463629 192.168.1.4 192.168.1.3 SNMP 141 report SNMP-USER-BASED-SM-MIB::usmStatsUnknownEngineIDs.0
4 2016-09-23 13:41:22.583987192 192.168.1.4 192.168.1.3 SNMP 141 report SNMP-USER-BASED-SM-MIB::usmStatsUnknownEngineIDs.0
5 2016-09-23 13:41:22.745228783 192.168.1.3 192.168.1.4 SNMP 180 get-request SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16974080
6 2016-09-23 13:41:22.745775804 192.168.1.3 192.168.1.4 SNMP 180 get-request SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16975104
7 2016-09-23 13:41:22.746921812 192.168.1.4 192.168.1.3 SNMP 157 report SNMP-USER-BASED-SM-MIB::usmStatsNotInTimeWindows.0
8 2016-09-23 13:41:22.747486766 192.168.1.3 192.168.1.4 SNMP 181 get-request SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16974080
9 2016-09-23 13:41:22.747921357 192.168.1.4 192.168.1.3 SNMP 157 report SNMP-USER-BASED-SM-MIB::usmStatsNotInTimeWindows.0
10 2016-09-23 13:41:22.748393948 192.168.1.3 192.168.1.4 SNMP 181 get-request SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16975104
11 2016-09-23 13:41:22.749550684 192.168.1.4 192.168.1.3 SNMP 185 get-response SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16974080
12 2016-09-23 13:41:22.750722158 192.168.1.4 192.168.1.3 SNMP 185 get-response SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16975104
Is the error org.snmp4j.MessageException : Operation not permitted is because of my multi-thread design or because of Linux environment firewall ?
Is it normal to have report SNMP-USER-BASED-SM-MIB::usmStatsNotInTimeWindows.0 at the first time or is because of my multi-thread design ?
So we I launch a lot of request like in the same execution program, I have at each time, the first time report SNMP-USER-BASED-SM-MIB::usmStatsNotInTimeWindows.0, and the other request/response are normal like this :
No. Time Source Destination Protocol Length Info
1 2016-09-23 13:32:50.405161233 192.168.1.3 192.168.1.4 SNMP 181 get-request SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16975104
2 2016-09-23 13:32:50.405629005 192.168.1.3 192.168.1.4 SNMP 181 get-request SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16974080
3 2016-09-23 13:32:50.406982591 192.168.1.4 192.168.1.3 SNMP 185 get-response SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16975104
4 2016-09-23 13:32:50.408058738 192.168.1.4 192.168.1.3 SNMP 185 get-response SNMPv2-SMI::enterprises.7483.2.2.4.3.4.190.1.3.16974080
Thank you for your help/opinion (multi-thread design problem or linux firewalling problem or else ) ?
Best Regards,
Thierry TESSALONIKOS
More information about the SNMP4J
mailing list