not a support question
Anonymous
Unregistered
I am designing a data model and am using a concatenation of three attributes as the row key, ordered respectively: date (hour grain) & app_name & client_uuid.
When doing queries or grabbing data for MapReduce jobs, I would like query by date, or possibly by date & app_name. client_uuid is there more for uniqueness and distribution of the hashing. Is there a way to handle this situation? Say, search by 'date & *' or search by 'date & app_name & *'.
Thanks in advance for the help!
Since RandomPartitioner distributes your row keys randomly throughout the cluster there is no way to do what you are proposing. One way to achieve what you want might be to keep a manual index that allows you to enumerate the keys you want to retrieve.
So for example In one column family have rows where the row keys are dates and the column values are app names for that given date. Then another column family where row keys are app names and the column values are client_uuids for that app_name. This would allow you to construct the row keys to retrieve all the rows for a given date or a given date and app name.
My specific example might not apply directly to your use case but the basic concept is probably what you need. There is no way to do range queries on keys so you will need to keep an index for finding all the keys you want to retrieve.
