Re: Possible hang in 10.1 with JSON query over partially indexed partitions - Mailing list pgsql-general

From Tom Lane
Subject Re: Possible hang in 10.1 with JSON query over partially indexed partitions
Date
Msg-id 7994.1516324038@sss.pgh.pa.us
Whole thread Raw
In response to Possible hang in 10.1 with JSON query over partially indexedpartitions  (Paul Jones <admin@cmicdo.com>)
List pgsql-general
Paul Jones <admin@cmicdo.com> writes:
> I may have discovered a situation in 10.1 where EXECUTEing a PREPARED
> statement acting on JSON data in partitioned tables hangs in an
> infinite loop for a particular set of data.  Unfortunately, the data is
> proprietary, so I did the best I could below to describe what happened.

You haven't provided any real reason to think that this represents an
infinite loop and not just a really slow query.  The query's filter
conditions are complex enough to render the planner's estimates basically
works of fiction :-(, so I have no faith in the estimates suggesting that
only a relatively small number of rows will get processed.  It's
definitely making some progress, because your successive backtraces
show different blocks getting read.

            regards, tom lane


pgsql-general by date:

Previous
From: Seamus Abshere
Date:
Subject: warning for subquery that references a table but not its columns
Next
From: Tom Lane
Date:
Subject: Re: warning for subquery that references a table but not its columns