cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jonathan Ellis <jbel...@gmail.com>
Subject Re: quorum / hinted handoff
Date Fri, 20 Nov 2009 22:54:06 GMT
Only writes to nodes that do not affect the consistencylevel can be
hinted, otherwise you've broken the contract that when you do quorum
reads after quorum writes you will always see the latest data.

So if you want to have writes available w/ one node down you can
either use RF=3 so quorum is only 2 (and the missing one will be
HH'd), or you can relax to CL.ONE.

-Jonathan

On Fri, Nov 20, 2009 at 4:47 PM, B. Todd Burruss <bburruss@real.com> wrote:
> but doesn't cassandra support hinted handoff?  so if a required node is
> unavailable on a write, then the next node in the ring will take the
> write with a "hint" of where to put it once the node is available again.
>
> if there isn't a "next node" then it will fail.  right?
>
> On Fri, 2009-11-20 at 14:25 -0800, Edmond Lau wrote:
>> The key is that an RF of 2 and a cluster size of 3 means that every
>> piece of data is owned by 2 nodes, and the assignment of data to the 2
>> nodes responsible for that data is determined by positions in the hash
>> ring, not by what's up and running.  So if you have nodes A, B, and C,
>> and a given key hashes to nodes A and B, then taking down either A or
>> B will prevent a quorum write to those two nodes from succeeding.
>>
>> Edmond
>>
>> On Fri, Nov 20, 2009 at 2:14 PM, B. Todd Burruss <bburruss@real.com> wrote:
>> > not really.  it seems that if i start with 3 nodes, remove 1 of them, i
>> > should still have a quorum, which is 2.  this is not what i experience.
>> >
>> > On Fri, 2009-11-20 at 16:03 -0600, Jonathan Ellis wrote:
>> >> Oh, okay.  Then it's working as expected.
>> >>
>> >> Does it make more sense to you now? :)
>> >>
>> >> -Jonathan
>> >>
>> >> On Fri, Nov 20, 2009 at 3:43 PM, B. Todd Burruss <bburruss@real.com>
wrote:
>> >> > this was on the build i got yesterday, 882359.
>> >> >
>> >> > ... and you are correct about if you start with 2 nodes and take one
>> >> > down - there isn't a quorum and the write/read fails.  i tested that
as
>> >> > well.
>> >> >
>> >> > thx!
>> >> >
>> >> >
>> >> > On Fri, 2009-11-20 at 15:30 -0600, Jonathan Ellis wrote:
>> >> >> On Fri, Nov 20, 2009 at 11:31 AM, B. Todd Burruss <bburruss@real.com>
wrote:
>> >> >> > one more point on this .. if i only start a cluster with 2
nodes, and i
>> >> >> > use the same config setup (RF=2, etc) .. it works fine.  it's
only when
>> >> >> > i start with the 3 nodes and remove 1.  in fact, i remove
the node
>> >> >> > before i do any reads or writes at all, completely fresh database.
>> >> >>
>> >> >> That sounds like a bug.  If you have 2 nodes, RF of 2, and take
one
>> >> >> node down then quorum anything should always fail.
>> >> >>
>> >> >> Is this on trunk still?
>> >> >>
>> >> >> -Jonathan
>> >> >
>> >> >
>> >> >
>> >
>> >
>> >
>
>
>

Mime
View raw message