Cassandra
  1. Cassandra
  2. CASSANDRA-1430

SSTable statistics causing intermittent CL test failures in trunk.

    Details

    • Type: Bug Bug
    • Status: Resolved
    • Priority: Major Major
    • Resolution: Fixed
    • Fix Version/s: 0.7 beta 2
    • Component/s: Core
    • Labels:
      None

      Description

      [junit] Testcase: testCleanup(org.apache.cassandra.db.CommitLogTest): FAILED
      [junit] 2 != 1
      [junit] junit.framework.AssertionFailedError: 2 != 1
      [junit] at org.apache.cassandra.db.CommitLogTest.testCleanup(CommitLogTest.java:69)
      [junit]
      [junit]

      I see this 1-2 times a day.

      1. 1430.txt
        22 kB
        Brandon Williams

        Activity

        Hide
        Hudson added a comment -

        Integrated in Cassandra #533 (See https://hudson.apache.org/hudson/job/Cassandra/533/)

        Show
        Hudson added a comment - Integrated in Cassandra #533 (See https://hudson.apache.org/hudson/job/Cassandra/533/ )
        Hide
        Jonathan Ellis added a comment -

        fixed by CASSANDRA-1471

        Show
        Jonathan Ellis added a comment - fixed by CASSANDRA-1471
        Hide
        Jonathan Ellis added a comment -

        this broke StreamingTransferTest (easy to miss in the ant output because it is a timeout rather than a failure)

        Show
        Jonathan Ellis added a comment - this broke StreamingTransferTest (easy to miss in the ant output because it is a timeout rather than a failure)
        Hide
        Brandon Williams added a comment -

        Committed.

        Show
        Brandon Williams added a comment - Committed.
        Hide
        Jonathan Ellis added a comment -

        +1

        Show
        Jonathan Ellis added a comment - +1
        Hide
        Brandon Williams added a comment -

        Oops, I'd forgotten about more tests being added due to CASSANDRA-1155's problems. Patch updated to fix tests.

        Show
        Brandon Williams added a comment - Oops, I'd forgotten about more tests being added due to CASSANDRA-1155 's problems. Patch updated to fix tests.
        Hide
        Jonathan Ellis added a comment -

        I'm getting build errors in a couple test classes

        Show
        Jonathan Ellis added a comment - I'm getting build errors in a couple test classes
        Hide
        Brandon Williams added a comment -

        Patch to remove sstable stats from the system table, and make them their own sstable component.

        Show
        Brandon Williams added a comment - Patch to remove sstable stats from the system table, and make them their own sstable component.
        Hide
        Jonathan Ellis added a comment -

        Let's just create an EstimatedHistogram serializer and write out one for each of the row sizes and column counts. SSTable.components will have to be updated.

        Show
        Jonathan Ellis added a comment - Let's just create an EstimatedHistogram serializer and write out one for each of the row sizes and column counts. SSTable.components will have to be updated.
        Hide
        Gary Dusbabek added a comment -

        I can't disagree. CASSANDRA-1382 should be on the list of things that get reverted when we figure out a better way of statistics.

        Show
        Gary Dusbabek added a comment - I can't disagree. CASSANDRA-1382 should be on the list of things that get reverted when we figure out a better way of statistics.
        Hide
        Jonathan Ellis added a comment -

        it already feels like a rather ugly set of workarounds, and maybe it's time to cut our losses

        Show
        Jonathan Ellis added a comment - it already feels like a rather ugly set of workarounds, and maybe it's time to cut our losses
        Hide
        Gary Dusbabek added a comment -

        >is it time to admit that storing sstable info, in sstables, wasn't a good idea?
        What if statistics operations ran on its own stage that could be halted at critical moments (flushing, for example).

        Show
        Gary Dusbabek added a comment - >is it time to admit that storing sstable info, in sstables, wasn't a good idea? What if statistics operations ran on its own stage that could be halted at critical moments (flushing, for example).
        Hide
        Jonathan Ellis added a comment -

        (this isn't just a test issue, i'm pretty sure it could bork drain, too)

        Show
        Jonathan Ellis added a comment - (this isn't just a test issue, i'm pretty sure it could bork drain, too)
        Hide
        Jonathan Ellis added a comment -

        this is caused by sstable statistics being written after the flush (see ssTableWriter.closeAndOpenReader)

        is it time to admit that storing sstable info, in sstables, wasn't a good idea?

        Show
        Jonathan Ellis added a comment - this is caused by sstable statistics being written after the flush (see ssTableWriter.closeAndOpenReader) is it time to admit that storing sstable info, in sstables, wasn't a good idea?
        Hide
        Gary Dusbabek added a comment -
        Show
        Gary Dusbabek added a comment - To prove that I'm not hallucinating: http://ci.apache.org/builders/cassandra-trunk/builds/437/steps/compile/logs/stdio

          People

          • Assignee:
            Brandon Williams
            Reporter:
            Gary Dusbabek
            Reviewer:
            Jonathan Ellis
          • Votes:
            0 Vote for this issue
            Watchers:
            1 Start watching this issue

            Dates

            • Created:
              Updated:
              Resolved:

              Development