Permalink
Switch branches/tags
Commits on Sep 24, 2013
  1. poor (hu)man's consensus for bucket types

    jrwest committed Sep 24, 2013
    Cluster Metadata's eventually consistency and bucket type semantics
    don't play well together. This commit addresses this disharmony by
    adding some hacky consensus. It does this in two steps. First, all
    writes to bucket types are now serialized through the claimant. Second,
    the status of the bucket type can be checked. See riak_core_bucket_type:status/1
    for more details. Using the status, a user can decide whether its "safe" to
    update the type (or determine if it has been successfully created on all nodes,etc).
    Many of the cases where updating a type is not recommended in the docs of status/1
    are safe to update only if the claimant remains alive (or goes down but comes back up
    w/o the user downing the node). However, this is error prone and brittle,
    so these cases are not recommended.
    
    One case not handled by this commit is if a user does indeed go against recommendations
    and downs a claimant when all recent bucket type's update have not reached the new claimant.
    The creation semantics in this case would be undefined.
Commits on Sep 23, 2013
  1. Merge pull request #393 from basho/jrw-hashtree-eqc-fix

    jrwest committed Sep 23, 2013
    update hashtree_eqc. get it passing on eqc 1.29.1
  2. Merge pull request #394 from basho/bugfix/use-branch-tg-in-rebar-config

    engelsanchez committed Sep 23, 2013
    Fix rebar.config to use {branch, "name"}
  3. Fix rebar.config to use {branch, "name"}

    engelsanchez committed Sep 23, 2013
    The rebar.config file currently is using "master" by itself, which does
    not make rebar actually update that branch on update-deps.
Commits on Sep 22, 2013
  1. update hashtree_eqc. get it passing on eqc 1.29.1

    jrwest committed Sep 22, 2013
    * get rid of make_treevars since the {call, erlang,'*', _} was the
    cause of unknown precondition failures and the same generation can be
    done without it
    * remove initial_state function and just pass the rest of the
    initial state to commands/2
    * use pretty_commands/4 so failures are easier to track down
  2. Merge branch 'jrw-cm-hashtree-rebased' into develop

    jrwest committed Sep 22, 2013
    Conflicts:
    	rebar.config
  3. link hashtree update proc to metadata_hashtree

    jrwest committed Sep 22, 2013
    otherwise the caller is not notified of its crash and waits
    infinitely
  4. prevent leakage of hsahtree_tree snaphots

    jrwest committed Sep 22, 2013
    delete the ets data if its still valid, catch the error otherwise
  5. hashtree_tree: better cleanup before creating node

    jrwest committed Sep 22, 2013
    extends hashtree:destroy to take a path and pass it directly to
    eleveldb
Commits on Sep 20, 2013
  1. Merge pull request #386 from basho/adt-security

    Vagabond committed Sep 20, 2013
    Add security to Riak
  2. Cleanup and refactor

    Vagabond committed Sep 20, 2013
  3. More work on security

    Vagabond committed Sep 16, 2013
  4. Remove extra newline

    Jared Morrow authored and Vagabond committed Jul 18, 2013
  5. Logging the plaintext password would be bad

    Jared Morrow authored and Vagabond committed Jul 18, 2013
  6. Remove unneeded case statement

    Jared Morrow authored and Vagabond committed Jul 18, 2013
  7. Updates to pw_auth and riak_core_security modules

    Jared Morrow authored and Vagabond committed Jul 18, 2013
  8. Move password auth into its own module

    Jared Morrow authored and Vagabond committed Jul 17, 2013
  9. Fix hmac algorithm

    Jared Morrow authored and Vagabond committed Jul 16, 2013
  10. Initial pass at generating a pbkdf2 pass

    Jared Morrow authored and Vagabond committed Jul 16, 2013
  11. Add pbkdf2 as a dep

    Jared Morrow authored and Vagabond committed Jul 16, 2013