This discussion is archived
1 2 Previous Next 20 Replies Latest reply: Jul 29, 2013 10:22 PM by BillyVerreynne Go to original post RSS
  • 15. Re: default service only connects to one node
    user13454469 Newbie
    Currently Being Moderated

    TESTDB =

      (DESCRIPTION =

      (ADDRESS_LIST =

      (LOAD_BALANCE = yes)

        (ADDRESS = (PROTOCOL = TCP)(HOST = scan-test)(PORT = 1527)))

        (CONNECT_DATA =

          (SERVER = DEDICATED)

          (SERVICE_NAME = TESTDB)

        )

      )

     

    Billy thanks for the info.  One other thing i tried was i used the above string(added the address_list and load_balance=yes), once i did that the session were going back and forth between the 2 nodes....but my question is the one original i had(below)...why was that not doing anything.... isint one of the purpose of the SCAN_TEST is to do the load balancing between the nodes even when i do not explicitly say load_balance=yes....

     

    TESTDB =

      (DESCRIPTION =

        (ADDRESS = (PROTOCOL = TCP)(HOST = scan-test)(PORT = 1527))

        (CONNECT_DATA =

          (SERVER = DEDICATED)

          (SERVICE_NAME = TESTDB)

        )

      )

     

    as far as sqlnet trace...would have the below entries be sufficient for the trace?

    SQLNET.EXPIRE_TIME=5

    TRACE_LEVEL_SERVER=16

    TRACE_LEVEL_CLIENT=16

    TRACE_TIMESTAMP_CLIENT=TRUE

    TRACE_UNIQUE_CLIENT=TRUE

    TRACE_DIRECTORY_CLIENT=C:\TRACE

    TRACE_DIRECTORY_SERVER=/backup/sql_trace/

    DIAG_ADR_ENABLED=OFF

  • 16. Re: default service only connects to one node
    BillyVerreynne Oracle ACE
    Currently Being Moderated

    Entries seem fine. I usually use the following myself:

    DIAG_ADR_ENABLED = off
    TRACE_LEVEL_CLIENT = admin
    TRACE_DIRECTORY_CLIENT = /home/billy/trace
    TRACE_UNIQUE_CLIENT = on
    
  • 17. Re: default service only connects to one node
    user13454469 Newbie
    Currently Being Moderated

    Billy Thanks for the info.

     

    We had restarted the cluster and ever since then it is behaving the way it suppose to be.  So i am not sure what was causing the issue.  Altough one thing i noticed after restarting the cluster...for about 15 mins straight i was pinging to scan-test from my laptop and i kept getting the below... 

     

    Pinging scan-test [110.20.10.81] with 32 bytes of data:     ----------->> one of our SCAN IP's

    Reply from 110.20.10.3: TTL expired in transit.                 ----------->> i believe one of the routers?

    Reply from 110.20.10.3: TTL expired in transit.

    Reply from 110.20.10.3: TTL expired in transit.

    Reply from 110.20.10.3: TTL expired in transit.

     

    Ping statistics for 10.26.17.81:

        Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),

     

    One thing to note, while i was doing the same from the server ping scan-test....it was giving me the reply back in the normal...only when i was doing it from my laptop i kept getting the TTL Expired in transit msg and obviously i couldnt connect to the DB either using the SCAN...but after those 15 mins passed, it was working just fine and now it doing its load balancing....

     

    Any ideas what that could be about?

  • 18. Re: default service only connects to one node
    BillyVerreynne Oracle ACE
    Currently Being Moderated

    From what you describe, it seems to me you have some kind of issue on your network.

     

    An ICMP packet starts of with a TTL (Time To Live) counter. Each hop (e.g. router) decrements the counter. When it reaches 0, TTL expires and that device responds to the ICMP. (basically how a traceroute works).

     

    TTL should normally not expired - unless it is set low via CLI switches for the ping command. Or the ICMP packet is bounced over more hops than usual.

     

    Run a traceroute (called tracert on Windows) to the SCAN IP from your laptop - should tell you how many hops there should be. For a corporate network, I would be surprised if it is 5+.

  • 19. Re: default service only connects to one node
    user13454469 Newbie
    Currently Being Moderated

    Hi Billy,

     

    It is actually doing 7 hops before reaching the scan destination.

     

    But i think one reason for 7 hops is because the server is sitting is NY and we are in Texas, and i am thinking that is probably why.  But i maybe wrong.  What can be done to trouble shoot the issue.  Like i mentioned, this ONLY happened when we shutdown the cluster(both the nodes) and then bring everything back up.  it takes about 20 mins before the ping responds(until then we get TTL expire)...

  • 20. Re: default service only connects to one node
    BillyVerreynne Oracle ACE
    Currently Being Moderated

    7 hops for such a distance is fine.

     

    I have never seen such a problem with Grid/RAC (been using it for many years). So my initial reaction is that this is some kind of networking issue. Either at o/s level on the RAC servers, or at the actual network infrastructure layer.

     

    Assuming that the actual boot and RAC startup do not itself take 20 minutes?

     

    The error sounds like a MAC change for an IP that takes some time to propagate via the switches - but just shooting from the hip...

     

    You need a network specialist to assist. Things to check would be switch ports (see what MACs are registered), arp and routing tables, and so on.

1 2 Previous Next

Legend

  • Correct Answers - 10 points
  • Helpful Answers - 5 points