Hi Does apache phoenix is supported for mapR version 5.0,5.1 and 5.2?
If yes which version of phoenix is supported for respective mapR versions?
If no then does mapR has plans to support apache phoenix in future?
Phoenix will never be supported by MapRDB.
Phoenix requires the use of HBase's Coprocessors and IIRC there were some changes to HBase as well.
IMHO while having server side extensions is needed, the implementation found in HBase creates the potential for security and stability issues. Running end user code inside the region servers is not a good thing. There is no isolation.
In addition, Phoenix is on a dead end in terms of evolution. There's an Apache project to offer SQL against multiple sources.
Thank for the valuable information Michael Segel
Can you please point to Apache project that you have mentioned in the reply.
Need to correct....
There's Apache ignite which is cross platform front end into an in-memory,
And then there's Tajo which looks to be a hive replacement but doesn't support Hbase or MapR yet.
There are so many Apace projects, I am not sure if I misspoke or just didn't see it.
However, why do you want to do SQL over Hbase/MapRDB? Neither are really meant for SQL.
Hi ASHISH DIWAN,
Please see Tug and Mufeed's answers on what MapR supports of using Phoenix:
phoenix query support on MapR 5.1
You are always welcome to follow Products and Services blogs to receive notifications on MapR latest new product releases.
I should have clarified that if you want to run HBase, you can and it is supported, however I really don't recommend running HBase when you have MapR-DB available.
HBase, depending on your administrator and how you are using it can be very unstable, especially if you're writing and using homegrown co-processors.
Thanks for your reply.
Summarizing your answers,
1. Phoenix is not supported by MaprDB and never will be supported.
2. Phoenix with Hbase is supported by mapR but can result into instability of the cluster.
Please do correct me on to these points.
Just a clarification.
Hbase won't impact the cluster, but only Hbase.
There are a couple of issues... if HBase isn't tuned properly, under load, it can fail where you have a cascading failure of region servers.
With co-processors, you can end up with end user or local developers writing a co-processor that runs on the same JVM as the region server so its possible to kill performance or take down the region server with bad code. Note: the was a fix that allowed you to disable co-processors or to disallow any after the masters are installed. (those set up in hbase-site.xml).
So your MapR-FS is up, Yarn is up, and other features are up... just that HBase can crash.
That's an issue w HBase and on the MapR platform, you're better off running MapR-DB.
Retrieving data ...