Skip to content
Commits on Oct 16, 2013
  1. @jaredmorrow
Commits on Oct 7, 2013
  1. @slfritchie

    Merge pull request #80 from basho/bugfix/bwf/fold-req-upgrade

    catch and upgrade riak_core_fold_req_v1
    slfritchie committed Oct 6, 2013
Commits on Oct 5, 2013
  1. catch and upgrade riak_core_fold_req_v1

    If this request isn't caught, it is dropped on the floor (with an "unknown
    message" log), and handoff freezes. This is related to basho/riak#407.
    Bryan Fink committed Oct 4, 2013
Commits on Oct 2, 2013
  1. @cmeiklejohn
  2. @cmeiklejohn
  3. add index to call to riak_core_ring:index_owner

    dialyzer complained about riak_core_ring:index_owner/1 not existing,
    because it's index_owner/2 and the Index argument was missing here
    Bryan Fink committed Oct 2, 2013
  4. remove incorrect dialyzer.ignore-warnings

    Bryan Fink committed Oct 2, 2013
Commits on Oct 1, 2013
  1. change if to case so dialyzer can understand it

    The logic before and after this commit should be exactly the same, but dialyzer
    reports the following for the `if` version:
    
    "riak_pipe_builder.erl:206: Clause guard cannot succeed."
    
    three out of four developers prefer the `case` version anyway
    Bryan Fink committed Oct 1, 2013
  2. correct unmatched return values

    dialyzer found many places where the return values of functions were ignored
    
    this commit adds failing matches or don't-care matches where appropriate
    Bryan Fink committed Oct 1, 2013
  3. fix type of riak_pipe_vnode:reply_to_blocker

    riak_core_vnode:reply mimics various erlang message sending functions, which means
    some return 'true', but others return the message that was sent, so the type for
    this function must be any()
    Bryan Fink committed Oct 1, 2013
  4. refactor riak_pipe_state:do_update to only return ok

    this resolves a dialyzer error, and also makes it clearer what uses the monitor
    Bryan Fink committed Oct 1, 2013
  5. simplify riak_pipe_builder_sup:builder_pids return value

    This removes the use of types not exported from 'supervisor' while also making
    the return value match the name of the function.
    Bryan Fink committed Oct 1, 2013
Commits on Aug 23, 2013
  1. @jtuple

    Setup develop branch

    jtuple committed Aug 23, 2013
Commits on Aug 13, 2013
  1. @jaredmorrow

    Merge branch '1.4'

    Conflicts:
    	rebar.config
    jaredmorrow committed Aug 13, 2013
Commits on Aug 1, 2013
  1. @rzezeski

    Update riak_core 1.4.1

    rzezeski committed Aug 1, 2013
Commits on Jul 11, 2013
  1. @jaredmorrow

    Merge branch '1.4'

    jaredmorrow committed Jul 11, 2013
Commits on Jun 26, 2013
  1. @jaredmorrow

    Roll version riak_pipe 1.4.0

    jaredmorrow committed Jun 25, 2013
Commits on Jun 25, 2013
  1. @jaredmorrow

    Merge branch '1.3'

    Conflicts:
    	rebar.config
    jaredmorrow committed Jun 25, 2013
Commits on May 30, 2013
  1. @jrwest

    Merge pull request #76 from basho/jrw-dynamic-ring

    limited support for ring resizing
    jrwest committed May 30, 2013
  2. @jrwest

    fix typespec

    jrwest committed May 30, 2013
Commits on May 21, 2013
  1. Merge pull request #75 from branch 'bwf-lower-log-level'

    Bryan Fink committed May 21, 2013
Commits on May 12, 2013
  1. @jaredmorrow

    Bump riak_core dep to 1.3.2

    jaredmorrow committed May 11, 2013
Commits on May 11, 2013
  1. @jaredmorrow

    Roll riak_pipe version 1.3.2

    jaredmorrow committed May 11, 2013
  2. @evanmcc

    Merge pull request #74 from basho/bwf-fitting-pulse-1.3

    Backport fix in issue 73 to 1.3 branch
    evanmcc committed May 10, 2013
Commits on May 9, 2013
  1. @jrwest

    limited support for ring resizing

    Provides ability for clusters using riak_pipe to use ring
    resizing. Ultimately, this is a hack that includes just enough support
    for clusters to successfully transition between the old and new ring
    without riak_pipe getting in the way. It does this by lying to
    riak_core's handoff subsystem during resize. The lie it tells is that
    even when it has outstanding work/queue items it will report that it
    is empty. This will allow the pipe vnode to continue processing work
    in the old ring since vnodes are not deleted and shutdown until the
    new ring is installed. This does, however, come with the drawback that
    work that is in-flight when the new ring is installed has undefined
    behaviour. The work may complete or may be terminated by the cleanup
    processes that follows the installation of the new ring.
    
    Better support for pipe has several challenges. Some can be easily
    addressed through extra support in riak_core but the fundamental
    challenge is that riak_pipe hands off worker archives. In a larger or
    smaller ring parts of those archrives would be owned by new
    partitions which would force those archives to be "re-opened" and
    disseminated during handoff.
    jrwest committed May 9, 2013
Commits on May 2, 2013
  1. lower "fitting was gone" log to debug level

    Issue #50 is not the only complaint about the message,
    "Pipe worker startup failed:fitting was gone before startup".
    It's not an important message, though. It *might* tell you that a
    pipe vnode is lagging a bit, but most often it's just noise about
    pipes that shut down for other reasons (MapReduce errors).
    
    This commit lowers the log level of the message to 'debug' and
    rewords it, since it really has nothing to do with the worker
    failing to start.
    Bryan Fink committed May 2, 2013
Commits on May 1, 2013
  1. cherry-pick commit 041cee5 into 1.3 branch

    It was desired to have the sink-sync-send crash fix backported to
    the 1.3 branch. This is a cherry-pick of the commit that fixed the
    issue from pull request #73. The tests tests that demonstrate the
    issue have not been backported, to save merge hassle.
    
    Original commit comment:
    
    catch sink fsm exits for all reasons
    
    The reduce_fitting_pulse test showed that other exit reasons might
    be seen here. For example, 'normal' can happen if the sink shuts
    down after the sync event has been delivered.
    
    This commit broadens the catch to prevent those other reasons from
    killing the current process, and also narrows it to make sure exit
    is bubbling from the sync_send_event, and nowhere else.
    Bryan Fink committed Apr 23, 2013
  2. Merge pull request #73 from branch 'bwf-fitting-pulse'

    Bryan Fink committed May 1, 2013
Commits on Apr 23, 2013
  1. doc & clean reduce_fitting_pulse

    Cleanups:
     - compile non-pulse parts even when PULSE is not defined,
       to allow sanity checking with PULSE disabled
     - use eqc:testing_time instead of eqc:numtests to better
       match the eunit timeout expectation
    Bryan Fink committed Apr 23, 2013
  2. catch sink fsm exits for all reasons

    The reduce_fitting_pulse test showed that other exit reasons might
    be seen here. For example, 'normal' can happen if the sink shuts
    down after the sync event has been delivered.
    
    This commit broadens the catch to prevent those other reasons from
    killing the current process, and also narrows it to make sure exit
    is bubbling from the sync_send_event, and nowhere else.
    Bryan Fink committed Apr 23, 2013
  3. add 'raw' sink time to fitting pulse test

    Checking to see whether this can reproduce issue 48 or 49. It didn't,
    but the coverage may be useful anyway.
    Bryan Fink committed Apr 23, 2013
Something went wrong with that request. Please try again.