Hi Aaron,
How many regions are there in the LINEITEM table? The fact that you
needed to bump the
hbase.mapreduce.bulkload.max.hfiles.perRegion.perFamily setting up to
48 suggests that the amount of data going into a single region of that
table is probably pretty large.
Along the same line, I beli
Gabriel,
Do you guys provide pre-split mechanisms (sampling of import/query data,
splitting policies, etc.) or does the admin have to determine the split points?
I guess that begs the question of how you would do a basic ETL operation in
Phoenix?
How would you do the following on a 100 gigs
Hi John,
You can actually pre-split a table when creating it, either by
specifying split points in the CREATE TABLE statement[1] or by using
salt buckets[2]. In my current use cases I always use salting, but
take a look at the salting documentation[2] for the pros and cons of
this.
Your approache
I am wondering why many of these interesting features are not listed in
official release notes. https://phoenix.apache.org/release_notes.html
Best,
Afshin
On Aug 12, 2016, at 1:25 PM, Ankit Singhal
mailto:an...@apache.org>> wrote:
Apache Phoenix enables OLTP and operational analytics for Hadoop
(-cc other lists)
Hi Afshin,
The release notes you referenced are more meant to alert users about any
issues in the new release that you may run into over previous releases.
"Release notes provide details on issues and their fixes which may have
an impact on prior Phoenix behavior"
- Josh
Gabriel,
Thanks for the response I appreciate it.
I struggle to understand how to use split points in the create statement.
(1) Creating a table with Split Points:
CREATE TABLE stats.prod_metrics ( host char(50) not null, created_date date not
null,
txn_count bigint CONSTRAINT pk PRIM
Looks like the SYSTEM.FUNCTION table is names with a reserved word. Is this a
known bug?
0: jdbc:phoenix:stl-colo-srv073.splicemachine> !tables
++--+-+---+--+++-+--+---
This is good feedback, Afshin. Thanks for letting us know. I've updated the
download page to provide a link to the new fixes/features. Would be great
if this link could be dynamic (i.e. always point to the release notes from
the last released version). Anyone know how to do this?
I've also updated
Aaron,
Looks like a permission issue?
org.apache.hadoop.hbase.security.access.SecureBulkLoadEndpoint threw
java.lang.IllegalStateException: Failed to get FileSystem instance
java.lang.IllegalStateException: Failed to get FileSystem instance
at
org.apache.hadoop.hbase.security.access.Sec
Maybe this will help?
http://phoenix.apache.org/bulk_dataload.html#Permissions_issues_when_uploading_HFiles
bq. I struggle to understand how to use split points in the create
statement.
You can't always use split points - it depends on your schema and the
knowledge you have about the data being l
10 matches
Mail list logo
