Riak Recap for Oct. 1-3

Mark Phillips mark at basho.com
Mon Oct 4 13:14:54 EDT 2010

Morning, Afternoon, Evening to all,

Here is a great recap from the past three days. We have upcoming
talks, blogs, some great Q & A and gists from #riak, and new community



Community Manager
Basho Technologies


Riak Recap for Oct. 1-3

1) Germain Maurice will be giving a Riak presentation at the upcoming
Open Source Developers Conference happening in Paris on October 9th
and 10th.

You can get more details on the talk and conference here --->

Merci, Germain!

2) Q --- How is the real hash value for replicas calculated based on
the "preflist"? (from Gerald1 via #riak)

    A --- The hash is calculated first, then the next subsequent N
partitions are chosen for the preflist.

3) @rress wrote up a short blog post entitled "Why do you think Riak
is a column-store?"

Post here ---> http://rrees.wordpress.com/2010/10/03/why-do-you-think-riak-is-a-column-store/

4) Q --- Is there riak-admin "leave" option ? to get note out of the
ring ? (from artkrz via #riak)

   A --- You can get a full list of the admin commands and what they
do on the wiki --->

5) Guest27178, seancribbs, technoweenie and jpartogi had a brief chat
about whether Riak can be used on "normal" servers and why having at
least three Riak nodes makes a lot of sense.

Gist here ---> http://gist.github.com/609996

6) Q --- If we want to search over a bucket using map-reduce, do we do
it in the map phase or reduce phase? (from jpartogi via #riak)

    A --- Aside from the performance considerations of doing a
full-bucket map-reduce, searching is a form of filtering, which should
be done in a map phase.

7) benblack and Gerald1 had a brief conversation in #riak that started
with "Anyone know how bitcask handles replication?"

Gist here ---> http://gist.github.com/610008

8) jpartogi, benblack, drev1 and siculars had a quick convo in #riak
about how/if Riak sorts data in a bucket by default.

Gist here ---> http://gist.github.com/610027

9) Q --- If the size of key index exceeds the amount of memory, how
does Bitcask handle it? (from Gerald1 via #riak)

    A --- Obviously when using Bitcask, memory usage is something you
should keep a close eye on. If you do happened to exceed RAM, it will
and when swap is gone it will crash with "out of memory" errors. To
protect against this, we generally advise you to have >2x RAM
available in your cluster than you expect the size of your keyspace to

If you haven't seen it already, this spreadsheet is a great tool when
you need to do some calculations and capacity planning --->

10) @siculars will be giving a talk in NYC tomorrow called "Adding
Riak to your NoSQL Bag of Tricks."

Talk and registration info here --->

Thanks, Alexander!

11) OJ (@TheColonial) has written two great posts focusing on Erlang
web development (using Webmachine, Riak and ErlyDTL). Rumor has it
that he is hard at work on the third piece, but, in the meantime, he
has posted the code he's using to BitBucket.

Repo here ---> https://bitbucket.org/OJ/csd/overview

Thanks, OJ!

More information about the riak-users mailing list