1. 07 Feb, 2017 4 commits
    • Ryan Senior's avatar
      (PDB-3108) Add producer-timestamp to the command "header" · eae9aa99
      Ryan Senior authored
      This commit adds producer-timestmap to the POST parameters of command
    • Ryan Senior's avatar
      (PDB-3108) Consolidate producer-timestamps in the terminus · 3e8ad4d6
      Ryan Senior authored
      This commit changes the terminus to ensure producer timestamps are
      created from a single place. Prior to this patch they were created in
      several places, including having the ability to be passed in from
      Puppet itself (not possible since producer-timestamp is something PDB
      created). This commit also switches to 3 digit accuracy on
      timestamps. Prior to this we were using different accuracy for
      catalogs/facts vs. reports, both of which were just truncated once it
      was received by the PDB backend.
    • Ryan Senior's avatar
      (PDB-3108) Add validation of HTTP request params in specs · 570e9d81
      Ryan Senior authored
      This commit adds some validation of the POST parameters for the
      terminus' spec tests
    • Ryan Senior's avatar
      (PDB-3108) Add producer-timestamp as a command header · b3beba09
      Ryan Senior authored
      This commit accepts the producer-timestamp as a POST parameter and
      threads that parameter down the stack into the enqueuing code. The
      producer-timestamp is required to enable "bash in place". Without it,
      we could potentially bash a newer catalogs with a retried (or delayed)
      older catalog. Getting the timestamp from the header means we can
      continue to delay parsing of the body of the command until it's
      It's possible to not have a producer-timestamp for a command. Examples
      of that would be on import or if commands were in the queue right
      before an upgrade. When there is no producer timestamp present, we
      should NOT allow that command to be bashed and instead should process
      the command normally (i.e. FIFO). This commit also includes the
      necessary logic to consider producer-timestamp when deciding to bash
      the command (i.e. mark as deleted, then when dequeued, it's removed).
  2. 06 Feb, 2017 1 commit
  3. 03 Feb, 2017 5 commits
  4. 02 Feb, 2017 1 commit
  5. 31 Jan, 2017 1 commit
  6. 30 Jan, 2017 1 commit
    • Ryan Senior's avatar
      Merge pull request #2170 from... · a876d649
      Ryan Senior authored
      Merge pull request #2170 from camlow325/bug/master/close-datasource-in-max-connection-jdbc-pool-tests
      (MAINT) Close datasource in max connection jdbc pool tests
  7. 28 Jan, 2017 1 commit
    • Jeremy Barlow's avatar
      (MAINT) Close datasource in max connection jdbc pool tests · 3832c436
      Jeremy Barlow authored
      Previously, the max connections jdbc pool-construction tests did not
      close the connection pool data source before falling out of scope.
      Depending upon timing, this could cause the tests to fail on
      attempting to drop temp databases which are still busy being used
      by the connection pool.
      This commit wraps usage of the data source in each of the max
      connections tests with `with-open` calls, allowing the connection pool
      to be closed off before any calls to drop the temporary databases occur
      (and, therefore, for the tests to complete successfully).
  8. 27 Jan, 2017 3 commits
  9. 26 Jan, 2017 3 commits
  10. 25 Jan, 2017 2 commits
  11. 19 Jan, 2017 3 commits
  12. 18 Jan, 2017 1 commit
    • Jeremy Barlow's avatar
      (MAINT) Upgrade to clj-parent 0.3.2 · c45a26e9
      Jeremy Barlow authored
      This commit upgrades the clj-parent dependency from 0.3.0 to 0.3.2 in
      order to pick up a fix from trapperkeeper-status 0.7.1 where gc/cpu
      metric collection could corrupt the status endpoint.
  13. 17 Jan, 2017 1 commit
  14. 13 Jan, 2017 3 commits
  15. 12 Jan, 2017 1 commit
  16. 10 Jan, 2017 4 commits
  17. 09 Jan, 2017 5 commits