Leena Gupta 2014-01-17, 00:17
Stephen Sprague 2014-01-17, 00:42
Leena Gupta 2014-01-17, 03:54
Stephen Sprague 2014-01-17, 06:05
Leena Gupta 2014-01-18, 02:05
Mohammad Islam 2014-01-18, 02:26
Good 'ole permissions, eh? A shame the error thrown by the datanuculeus
code didn't expose this important piece of information.
Instead we get a red herring about "autocreate flag" nonsense and waste
hours of time. Egads.
+1 for sharing back!
On Fri, Jan 17, 2014 at 6:26 PM, Mohammad Islam <[EMAIL PROTECTED]> wrote:
> I never used postgres with Hive.
> These links might be helpful:
> Look for *Using Postgres for the Hive Metastore*
> *Configure the Metastore Service to Communicate with the PostgreSQL
> Remember these instructions are vendor specific.
> On Friday, January 17, 2014 6:06 PM, Leena Gupta <[EMAIL PROTECTED]>
> After several tries, I finally figured it out. Thought I'll post what I
> had to do in case others hit this problem.
> From what I observed, it looks like the Datanucleus error is related to
> the absence of proper access permissions to hiveuser in postgres. I
> uninstalled both hive and postgres, installed hive 0.10 and postgres 9.1.
> Then created the hive schema in postgres and granted permissions on all
> objects within the public schema to hiveuser.
> After this when I ran create table command, it again threw an error
> related to the table 'COLUMNS_V2'. So then I manually granted
> select,Insert,update delete for this table to hiveuser and now everything
> works fine :)
> On Thu, Jan 16, 2014 at 10:05 PM, Stephen Sprague <[EMAIL PROTECTED]>wrote:
> okay. so SEQUENCE_TABLE does indeed exist. that's the first thing to get
> out of the way then. hmm.
> I suspect these other tables are just artifacts from the differences in
> the hive versions.
> yeah, i'm not sure where to go from here and I couldn't find much via
> googling either.
> would it be possible to install hive v0.12? I can at least vouch for that
> version. (i'm using postgres v9.1 however.)
> Otherwise, we'll have to wait for someone else to chase this bugger down -
> which may not be a bad thing. :)
> On Thu, Jan 16, 2014 at 7:54 PM, Leena Gupta <[EMAIL PROTECTED]>wrote:
> Thanks for responding Stephen.
> I checked the tables in Postgres and the SEQUENCE_TABLE exists. However in
> comparison to the list you gave for Hive 0.12, the following tables are
> missing, not sure if these could be the cause of the datanucleus error :
> On Thu, Jan 16, 2014 at 4:42 PM, Stephen Sprague <[EMAIL PROTECTED]>wrote:
> On Thu, Jan 16, 2014 at 4:17 PM, Leena Gupta <[EMAIL PROTECTED]>wrote:
> Could not create "increment"/"table" value-generation container
> "SEQUENCE_TABLE" since autoCreate flags do not allow it.
> Interestingly enough this exact same question is posted here:
> and crickets for responses. the twist is is that its for mysql! which
> leads us to the fact that its a middleware (datanucleus?) problem - not a
> metastore vendor problem.
> for starters first thing i'd do (regardless of the autocreate flags in
> hive-site.xml) is to confirm that table (SEQUENCE_TABLE) exists in your
> postgres metastore database. "\dt" would list the tables. Is it present?
> PS this is what have for hive v0.12 on postgres if its any help.
> dwr_prod=> \dt
> List of relations
> Schema | Name | Type | Owner
> public | BUCKETING_COLS | table | postgres
> public | CDS | table | postgres
> public | COLUMNS_OLD | table | postgres
> public | COLUMNS_V2 | table | postgres
> public | DATABASE_PARAMS | table | postgres