Re: Creating A GIN index on JSONB column (large database)

From: Ron <ronljohnsonjr(at)gmail(dot)com>
To: pgsql-general(at)lists(dot)postgresql(dot)org
Subject: Re: Creating A GIN index on JSONB column (large database)
Date: 2022-08-09 16:53:28
Message-ID: eeabb0d8-a262-9f93-72b7-0952b0d406b8@gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

800GB/20TB = 4%.

You've got a 20TB database, and run out of space when *4% is added*.

Either you're running way to close to the edge, or I misunderstand something.

Either way, enable auto-scaling.
https://aws.amazon.com/about-aws/whats-new/2019/06/rds-storage-auto-scaling/

On 8/9/22 10:29, Taylor Smith wrote:
> Yes, the database was always going to be a minimum of 20TB however we
> build the indexes on the partition which at max is about 1.3TB, which in
> temp files will use about 800GB temporary files. Note it doesn't have the
> same effect when doing non-concurrently. I don't think rolling off would
> help as its likely largest partition will either be the current month or
> the previous month so the problem would still persist.
>
>
>
> On Tue, Aug 9, 2022 at 3:50 PM Rob Sargent <robjsargent(at)gmail(dot)com> wrote:
>
>
>
> > On Aug 9, 2022, at 7:45 AM, Taylor Smith <taylorjrsmith(at)gmail(dot)com>
> wrote:
> >
> > Thanks Rob,
> >
> > This is what I was thinking. Would you say it presents any risk then
> having a database unable to rebuild its own indexes if needed?
> >
> >
>
> No I would not.  The scale-up + rebuild should tell you how much you
> actually need and if you’ve over done the scale up you can scale down
> to what you need.  Was the original spec expecting 20+T?  Any chance
> of rolling older months off to some other node?
>

--
Angular momentum makes the world go 'round.

In response to

Browse pgsql-general by date

  From Date Subject
Next Message Daulat 2022-08-10 04:57:46 index row size 2720 exceeds btree version 4
Previous Message Ron 2022-08-09 16:46:03 Re: Creating A GIN index on JSONB column (large database)