cassandra-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeremy Hanna <>
Subject Re: plugins/triggers/coprocessors
Date Wed, 16 Feb 2011 15:25:43 GMT
Just wanted to let people who follow the user list know that if there is interest in something
like plugins, triggers, or coprocessors on the server-side with Cassandra, the ticket to follow
or get involved with (code, comments, etc) is CASSANDRA-1311:

On Feb 11, 2011, at 1:31 PM, Jeremy Hanna wrote:

> So from here I guess it's a matter of working out the comments/concerns presented on
1311 and any future discussion sounds like it belongs there.
> Like I said, I just wanted to initiate discussion since it had been a while and the dust
from 0.7 had settled.  It seems like an incredibly useful concept to have as a core feature.
 Another motivation was Ben Black presenting at Strata.  He had mentioned that he and Cliff
had worked through doing server side operations which sounded similar (though their effort
sounded like it was not generalizable).  I've talked to others in the community that have
hoped for a feature like this too.  In any case, since it crossed ticket boundaries, I thought
it would be most appropriate to gauge interest as a discussion thread.
> Hopefully this will help for people who would like to either help out with implementation
or give feedback as to how it can be made general purpose or more Cassandra-y.
> On Feb 11, 2011, at 1:11 PM, Jeff Hodges wrote:
>> As the dude that worked on the 1016 prototype, I agree with this.
>> On Fri, Feb 11, 2011 at 10:53 AM, Stu Hood <> wrote:
>>> Honestly, I think we should just mark 1016 a dupe and move forward with
>>> 1311: we won't be hurting anyone's feelings, and the implementation from
>>> 1016 is: 1. much, much less complete, 2. abandoned.
>>> On Fri, Feb 11, 2011 at 9:23 AM, Jeremy Hanna <>wrote:
>>>> Thanks Maxim - I'll just go ahead and BCC you and Hentschel and move the
>>>> discussion to the dev list.
>>>> Based on the comments on 1311 - did you have anything else to add to that
>>>> could we unify around 1016 or 1311 and work on getting that to a general
>>>> state of acceptance?  Were there any that were able to do some work on
>>>> either these days?  Or are we not at that point?
>>>> On Feb 11, 2011, at 10:36 AM, Maxim Grinev wrote:
>>>>> Hi all,
>>>>> Jeremy, thanks for starting the discussion! We don't follow the dev list
>>>> closely so it was a good idea to email it directly.
>>>>> It really seems to be about the same. To unify the discussions, we
>>>> propose to list the features of each solution and compare them feature by
>>>> feature. Here is the feature list for triggers:
>>>>>      • Triggers are set on a column family. Triggers are executed for
>>>> each mutation to the column family and parametrized by the mutation.
>>>>>      • The mutation, which is the trigger parameter, is the "new" value.
>>>> The trigger cannot see the "old" value.
>>>>>      • Triggers are executed asynchronously some time after the write
>>>> which fired it is acknowledged to the client.
>>>>>      • Triggers are executed once during normal execution. We guarantee
>>>> "at least once" execution in case of node failures.
>>>>> Cheers,
>>>>> Maxim
>>>>> On Thu, Feb 10, 2011 at 8:45 AM, Jeremy Hanna
>>>>> <> wrote:
>>>>>> Hey guys,
>>>>>> I was just wondering if it would be a good time to unify discussions
>>>> plugins/triggers/coprocessors?
>>>>>> I was going to email the dev list but since I don't know if all of
>>>> follow the dev list and you guys are the ones that expressed the most
>>>> interest, I thought I would start here.
>>>>> Yeah, they're all tackling basically the same problem. For which we
>>>>> should have a single solution.
>>>>> -ryan

View raw message