HBase Bulk Ingest Problems

HBase Bulk Ingest Problems

Issues that come up in HBase bulk ingest that should either be documented or handled in code or a FAQ

1. Unabled to load Snappy error...you'll basically get the error "java.lang.RuntimeException: native snappy library not available: this version of libhadoop was built without snappy support."

fix is to
export LD_LIBRARY_PATH=/usr/lib/hadoop/lib/native/
2. java.io.Exception: Exception in rename
Caused by Permission Denied: user=hbase, access=WRITE, inode="/tmp/foobar/id/d/":ec2-user:hadoop:drwr-xr-x

Seems that hbase user needs to be able to have write perms so it can rename/move the files. Should probably check for permissions with the hadoop API at the start of the "bulk-load" command and have our first bulk load ingest command set the perms permissively. e

3. It appears that the Hfile version may not be set to v3. I'm seeing:
   WARN [o.a.h.h.i.f.HFileWriterV2] A minimum of HFile version 3 is required to support cell attributes/tags. Consider setting hfile.format.version accordingly
   but i don't know if this is real...

4. Validate flag doesn't work (needs to either default to false or have explicit arity = 1). Probably could make it 'skip-validation' and default to false

Original JIRA Issue: https://geomesa.atlassian.net/browse/GEOMESA-2276

Key: GEOMESA-2276
Type: Improvement
Priority: Major
Status: To Do
Resolution: Unresolved
Reporter: Andrew Hulbert
Created: Tue, 1 May 2018 21:23:38 -0400
Updated: Fri, 17 Apr 2020 11:46:01 -0400

Comments:

557058:48912df4-0004-4f90-9915-db26684ec006 (Thu, 16 Apr 2020 17:56:10 -0400):

Austin Heyne Emilio Lahr-Vivaz Andrew Hulbert is this still an issue worth discussing?

557058:da7f6dfa-f0a1-468b-aec5-2cb61273feb5 (Fri, 17 Apr 2020 09:07:06 -0400):

I think we fixed 3 and 4, at least.

557058:3f10fbb6-8894-45e8-828a-0f6aa4281f1d (Fri, 17 Apr 2020 11:15:57 -0400):

1 is still a problem.

557058:48912df4-0004-4f90-9915-db26684ec006 (Fri, 17 Apr 2020 11:46:01 -0400):

Austin Heyne is #1 a GeoMesa problem or a problem with a given deployment / environment?

Start at the bulk-load command and reproduce the native Snappy error described in the issue, checking whether it is specific to GeoMesa or to the deployment environment. Review the listed HBase and Hadoop permission details, then document or handle the confirmed problem and verify that bulk ingest completes successfully.