When using the stickynotes app, instead of curl, the notes bucket is marked as being handled by the ‘notes’ module. When objects are stored in the notes bucket, functions from this module are called, and it is one of those functions that adds the reverse link to the groups object.

If you wanted to do the same with just curl, you’d need to execute something like:

> curl -X PUT -H “content-type: application/json” http://127.0.0.1:8098/jiak/notes –data “{\”bucket_mod\”:\”notes\”}”

Assuming you’re using the stickynotes riak config file, and therefore have the stickynotes ebin in your code path, and the notes and groups atoms exist, riak will make the calls you expect, and notes:affect_write/4 will add that reverse link for you.

HOWEVER, the curl command described required that you pull the latest riak (as of this evening). I just patched a bug while writing this reply. If it’s inconvenient for you to update riak, you can get the same effect by typing the following two lines at the erlang shell running riak (assuming you used the debug-fresh.sh script):

> {ok, C} = riak:local_client().
> C:set_bucket(<>, [{bucket_mod, notes}]).

“…I can now use the Jiak utility jaywalker to get all of the notes in the todo group with a single query:
$ curl http://127.0.0.1:8098/jiak/groups/todos/notes,_,_”

But at this point, I received an empty “results” list, as I’ve shown, instead of receiving “all of the notes” as it is stated. Did I misunderstand this sentence?

Your first curl command creates the “todos” item in the “groups” bucket.

Your second curl command creates the “blog” item in the “notes” bucket. You’ve added a link in the “notes/blog” item that points to the “groups/todos” item.

Your third curl command asks Riak to start at the “groups/todos” item and follow all “notes” links it has. The “groups/todos” item has no links at all, so Riak finds nothing to give you.

Your final curl command asks Riak to start at the “notes/blog” item and follow all “groups” links it has. It has one link (the one you added in the second curl command), so it follows that one, finds the “groups/todos” item and returns it to you.

Make sense?

$ sudo ./start-fresh.sh config/riak-demo.erlenv
$ curl -X PUT -H “Content-type: application/json” http://127.0.0.1:8098/jiak/groups/todos -d “{\”bucket\”:\”groups\”,\”key\”:\”todos\”,\”object\”:{\”name\”:\”todo\”},\”links\”:[]}”
$ curl -X PUT -H “Content-type: application/json” http://127.0.0.1:8098/jiak/notes/blog -d “{\”bucket\”:\”notes\”,\”key\”:\”blog\”,\”object\”:{\”text\”:\”finish blog post\”,\”x\”:0,\”y\”:0,\”z\”:0,\”color\”:\”green\”},\”links\”:[[\"groups\",\"todos\",\"open\"]]}”
$ curl http://127.0.0.1:8098/jiak/groups/todos/notes,_,_
{“results”:[[]]}
$ curl http://127.0.0.1:8098/jiak/notes/blog/groups,_,_
{“results”:[[{"object":{"name":"todo"},"vclock":"a85hYGBgzGDKBVIsjGWr2jOYEhnzWBk4EqKO8GUBAA==","lastmod":"Mon, 26 Oct 2009 18:01:44 GMT","vtag":"23zAcKJCpRQzn8HKxdy5jh","bucket":"groups","key":"todos","links":[]}]]}

Based on your instructions and from my understanding of links the results for the 4th and 5th commands seem to be switched. Did I miss something?

I still miss those ordinary 1,2,3… ids like the known @@identity or SEQUENCE, but I admit that they might be non-adequate to the volatile networking multi-node environment with no master node in it.

There is currently nothing in the dets_backend watching for a potential overflow, but limit-checking code could be added.

…or you could choose among the other available backends.

The last straw of fear before the thing is that under improbable conditions one of vnodes gets overflown way earlier than the others. Are there any triggers implemented or in mind to face that? Or it is totally unlikely?

BUT, you must realize that Riak, by its distributed nature, mostly skirts the space limitations of dets.

Each Riak vnode using the dets_backend will open its own table. The number of vnodes is determined by the ring_creation_size parameter in the configuration file. This is often set to something like 16 in developer setups, or 1024 or more in production deployments. 16*2GB or 1024*2GB is a different ballgame.

Furthermore, it is *not* a requirement of Riak that the entire dataset fit on one node. In fact, data stored with a N-value less than the number of nodes in your cluster will not be stored on all nodes. Therefore, the “maximum capacity” of a given cluster configuration is not as simple as the “maximum capacity” of any given node or backend.

My question was rather about that later moment. When all nodes are back up again or connection gets restored – I strongly believe the consistensy will be back for deleted keys as well. The test (experiment) will certainly demonstrate it.
And you are right – the reset won't need to delete the bucket, clearing all keys will suffice for my purposes.

The last (is there such thing?) question: is dets_backend capable to hold more than that notorious 2Gb? This one seems critical in the persistent storage selection and I could not find any definite answer yet.

Thank you.