Uploaded image for project: 'Hadoop Common'
  1. Hadoop Common
  2. HADOOP-11187

NameNode - KMS communication fails after a long period of inactivity

    Details

    • Type: Bug
    • Status: Closed
    • Priority: Major
    • Resolution: Fixed
    • Affects Version/s: 2.6.0
    • Fix Version/s: 2.7.0
    • Component/s: None
    • Labels:
      None
    • Target Version/s:
    • Hadoop Flags:
      Reviewed

      Description

      As reported by Aaron T. Myers :

      The issue is due to the authentication token that the NN has to talk to the KMS is expiring, AND the signature secret provider in the KMS authentication filter is discarding the old secret after 2x the authentication token validity period.
      If the token being supplied is under 1x the validity lifetime then the token will authenticate just fine. If the token being supplied is between 1x-2x the validity lifetime, then the token can be validated but it will be expired, so a 401 will be returned to the client and it will get a new token. But if the token being supplied is greater than 2x the validity lifetime, then the KMS authentication filter will not even be able to validate the token, and will return a 403, which will cause the client to not retry authentication to the KMS.

      The KMSClientProvider needs to be modified to retry authentication even in the above case

        Attachments

        1. HADOOP-11187.1.patch
          14 kB
          Arun Suresh
        2. HADOOP-11187.2.patch
          16 kB
          Arun Suresh

          Activity

            People

            • Assignee:
              asuresh Arun Suresh
              Reporter:
              asuresh Arun Suresh
            • Votes:
              0 Vote for this issue
              Watchers:
              10 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved: