Introduction of HBase namespaces into a pre-existing application?

[#1]

I've inherited the support role for an application that writes all of its tables to the default namespace in HBase. I'd like to be able to leverage namespaces so that I can manage a centralized instance of HBase that many instances of our application can then be configured/reconfigured to use.

Questions

• I understand how to use namespaces but was wondering if there was a easy way to facilitate this through HBase/Hadoop on the backend?
• What's required to use namespaces from our application's standpoint?
• Can our application simply do a "use <namespace>" and from that point be referencing a specific namespace's tables?

[#2 · Accepted solution]

@Sam Mingolelli

namespaces have been available since 0.96 and not very well documented. Here's a scenario I was able to dig up to attempt answering your question.

Step 1: Superuser (e.g. user hbase) creates namespace foo.

create_namespace ‘foo’

Step 2: Admin gives dba-bar full permissions to the namespace:

grant ’dba-bar', 'RWXCA', '@foo’ 

Note: namespaces are prefixed by @.

Step 3: dba-bar creates tables within the namespace:

create ’foo:t1', 'f1’ 

Step 4: dba-bar hands out permissions to the tables:

grant ‘user-x’, ‘RWXCA’, ‘foo:t1’ 

Note: All users will be able to see namespaces and tables within namespaces, but not the data.

The next best source of information would be the umbrela Jira for namespaces https://issues.apache.org/jira/browse/HBASE-8015. It has a design document in PDF form https://issues.apache.org/jira/secure/attachment/12580245/Namespace%20Design.pdf

Outside of that, please feel free to read the namespace source code or unit tests https://github.com/apache/hbase/blob/master/hbase-server/src/test/java/org/apache/hadoop/hbase/TestN.... Bottom line, to use namespaces see above, to make your current application compliant with namespaces, change the code to point to the tables within namespaces, I already suggested to use Clone Snapshot and Export Snapshot tools or write Mapreduce to export/import into new table within a new namespace. I will escalate this to our docs team to have a better document as it's obvious people are starting to use this. I would've pointed you to our docs if we had that info, that's why I sent you stack overflow link as this question has not come up before. Finally, if at some point you will want to grant authorization to namespaces using Ranger, here's a thread for that https://community.hortonworks.com/questions/17764/ranger-hbase-namespace.html.

[#3]

@Enis

@Devaraj Das @Josh Elser

[#4]

@Artem Ervits

Thank you, this is a much more substantial answer to what I was looking for. In reviewing this material and in my previous researching it does not look like there's any method available via the hbase-site.xml file or top level command where you can specify a default namespace wrt client calls. So our only recourse looks to be to modify our application so that it explicitly calls out <ns>.<table> instead of what it's doing now. I was hopeful there was something along the lines of a `use <ns>` type of operation that I could utilize to "pin" our application's calls to HBase tables to a specific namespace, but that doesn't seem to be the case. At any rate I appreciate your time and look forward to some more thorough docs around namespaces.

[#5]

Yes, HBase does not have a "use namespace" equivalent. I doubt that we will ever add one, since in most of the cases, the default namespace is the special namespace "default". However, changing an application to use a table in a namespace should be pretty trivial. Make sure that you are using TableName.valueOf() methods properly.

[#6]

Hi @Sam Mingolelli, namespaces are indeed poorly documented. Here is a list of hbase shell commands you can use:

alter_namespace, create_namespace, describe_namespace, 
drop_namespace, list_namespace, list_namespace_tables

The first 4 are self-explanatory. With list_namespace you can list all available namespaces. The system namespace is now called "hbase" and tables without namespace all go to the "default" namespace. With "list_namespace_tables <namespace>" you can list tables in a given namespace. After that you can reference created tables by 'namespace:table', note that a semicolon is used as the delimiter, not a dot like in the design document.