Traffic Server
  1. Traffic Server
  2. TS-436

Documentation for variable hardware sector size support

    Details

    • Type: Improvement Improvement
    • Status: Closed
    • Priority: Major Major
    • Resolution: Fixed
    • Affects Version/s: 2.1.6
    • Fix Version/s: Docs
    • Component/s: Documentation
    • Labels:

      Description

      TS-43

      Adds support for auto-detecting drives which have 4096 (well, between 512-8192) sector size
      and using that as the element of atomicity in the cache.

      Add (optional/hidden) config:

      CONFIG proxy.config.cache.force_sector_size INT

      to force a sector size (e.g. 4096) for ALL disks.

      SSDs and "advanced format" drives lie about their sector size and
      claim 512. The handling of this is still in flux in linuxland hence the option
      to force a particular sector size. Note: it is safe to force a higher size
      than the hardware supports natively as we still only count on atomicity
      in 512 byte increments.

      4096 sector size drives formatted for windows will have partitions aligned on 63 512
      byte sector boundaries which will mean they will be unaligned.. There are workarounds, but
      you need to do some research on your particular drive. For example, some drives have a one
      time option to switch the partition boundary, others might require reformatting or repartitioning.

      To be safe in Linux you could just use the entire
      drive: /dev/sdb instead of /dev/sdb1 and TS will do the right thing.

      This feature should work fine when we get out of the difficult transition stage,
      in the mean time caveat emptor.

      Partitions formatted to support hardware sector
      size of more than 512 (e.g. 4096) will result in all objects stored
      in the cache to be integral multiples of 4096 bytes which will result in
      some waste for small files.

        Activity

        Hide
        John Plevyak added a comment -

        I have also added code to auto-detect misaligned partitions on Linux so that if
        (for example) /sys/block/sda/sda1/alignment_offset is non-zero, ATS will offset
        reads/writes to that disk by that alignment. So if you have a misaligned
        partition, and if linux knows about it, ATS will compensate.

        Show
        John Plevyak added a comment - I have also added code to auto-detect misaligned partitions on Linux so that if (for example) /sys/block/sda/sda1/alignment_offset is non-zero, ATS will offset reads/writes to that disk by that alignment. So if you have a misaligned partition, and if linux knows about it, ATS will compensate.
        Hide
        Leif Hedstrom added a comment -

        Moving these all to a "Doc 3.x" release version.

        Show
        Leif Hedstrom added a comment - Moving these all to a "Doc 3.x" release version.
        Hide
        Daniel Gruno added a comment -

        Description of it has been added to http://trafficserver.apache.org/docs/trunk/admin/configuration-files/records.config (and since fixed, since it was initially wrong).
        Does it need more elaboaration, or will people get the gist of it? We could add the long description supplied in this ticket to it, but that might seem a bit out of place compared to the other config field descriptions. Thoughts, suggestions?

        Show
        Daniel Gruno added a comment - Description of it has been added to http://trafficserver.apache.org/docs/trunk/admin/configuration-files/records.config (and since fixed, since it was initially wrong). Does it need more elaboaration, or will people get the gist of it? We could add the long description supplied in this ticket to it, but that might seem a bit out of place compared to the other config field descriptions. Thoughts, suggestions?
        Hide
        Leif Hedstrom added a comment -

        Changing to Doc 3.4, hopefully someone picks these up .

        Show
        Leif Hedstrom added a comment - Changing to Doc 3.4, hopefully someone picks these up .
        Hide
        ASF GitHub Bot added a comment -

        GitHub user cjqian opened a pull request:

        https://github.com/apache/trafficserver/pull/219

        TS-436 Added documentation for hardware sector size support

        This resolves issue #TS-436. Added documentation re: auto-detecting of hardware sector size (4096 default).

        You can merge this pull request into a Git repository by running:

        $ git pull https://github.com/cjqian/trafficserver issues/TS-436

        Alternatively you can review and apply these changes as the patch at:

        https://github.com/apache/trafficserver/pull/219.patch

        To close this pull request, make a commit to your master/trunk branch
        with (at least) the following in the commit message:

        This closes #219


        commit 80e5e0409429f33a4ebeaa00016c3dc8cc7e4f8f
        Author: Crystal Qian <crystal_qian@cable.comcast.net>
        Date: 2015-06-11T17:12:00Z

        TS-436 Added documentation for hardware sector size support


        Show
        ASF GitHub Bot added a comment - GitHub user cjqian opened a pull request: https://github.com/apache/trafficserver/pull/219 TS-436 Added documentation for hardware sector size support This resolves issue # TS-436 . Added documentation re: auto-detecting of hardware sector size (4096 default). You can merge this pull request into a Git repository by running: $ git pull https://github.com/cjqian/trafficserver issues/ TS-436 Alternatively you can review and apply these changes as the patch at: https://github.com/apache/trafficserver/pull/219.patch To close this pull request, make a commit to your master/trunk branch with (at least) the following in the commit message: This closes #219 commit 80e5e0409429f33a4ebeaa00016c3dc8cc7e4f8f Author: Crystal Qian <crystal_qian@cable.comcast.net> Date: 2015-06-11T17:12:00Z TS-436 Added documentation for hardware sector size support
        Hide
        Crystal Qian added a comment -

        Why was I unassigned?

        Show
        Crystal Qian added a comment - Why was I unassigned?
        Hide
        ASF subversion and git services added a comment -

        Commit 620e61dc44e08a8ee0cbc08a0b2c668f4ad63d13 in trafficserver's branch refs/heads/master from Crystal Qian
        [ https://git-wip-us.apache.org/repos/asf?p=trafficserver.git;h=620e61d ]

        TS-436: Added documentation for hardware sector size support

        Show
        ASF subversion and git services added a comment - Commit 620e61dc44e08a8ee0cbc08a0b2c668f4ad63d13 in trafficserver's branch refs/heads/master from Crystal Qian [ https://git-wip-us.apache.org/repos/asf?p=trafficserver.git;h=620e61d ] TS-436 : Added documentation for hardware sector size support
        Hide
        ASF subversion and git services added a comment -

        Commit 5eaea5104e4a937085911dc0beb310446c832a4a in trafficserver's branch refs/heads/master from Phil Sorber
        [ https://git-wip-us.apache.org/repos/asf?p=trafficserver.git;h=5eaea51 ]

        TS-436: Add to CHANGES, fix line length, and remove trailing space

        This closes #219

        Show
        ASF subversion and git services added a comment - Commit 5eaea5104e4a937085911dc0beb310446c832a4a in trafficserver's branch refs/heads/master from Phil Sorber [ https://git-wip-us.apache.org/repos/asf?p=trafficserver.git;h=5eaea51 ] TS-436 : Add to CHANGES, fix line length, and remove trailing space This closes #219
        Hide
        ASF GitHub Bot added a comment -

        Github user asfgit closed the pull request at:

        https://github.com/apache/trafficserver/pull/219

        Show
        ASF GitHub Bot added a comment - Github user asfgit closed the pull request at: https://github.com/apache/trafficserver/pull/219
        Hide
        Phil Sorber added a comment -

        Crystal Qian, The assignee is changed when someone else is working on the ticket. It is not a commentary on your work.

        I have merged this patch.

        Thanks.

        Show
        Phil Sorber added a comment - Crystal Qian , The assignee is changed when someone else is working on the ticket. It is not a commentary on your work. I have merged this patch. Thanks.

          People

          • Assignee:
            Phil Sorber
            Reporter:
            John Plevyak
          • Votes:
            0 Vote for this issue
            Watchers:
            6 Start watching this issue

            Dates

            • Created:
              Updated:
              Resolved:

              Development