public inbox for [email protected]
help / color / mirror / Atom feedFrom: Ron Johnson <[email protected]>
To: pgsql-admin <[email protected]>
Subject: Re: Adding New Column with default value.
Date: Mon, 28 Apr 2025 22:06:21 -0400
Message-ID: <CANzqJaBdQ5CsvXzDNWd19-auoLyMT0V7UHjKvXd8Bp4T0oeNZg@mail.gmail.com> (raw)
In-Reply-To: <[email protected]>
References: <CAHOGQfVmfJ5rpxfMSqJjTZGqN5ss-hdsCdTHsrxHAi2mUwWwmw@mail.gmail.com>
<[email protected]>
On Mon, Apr 28, 2025 at 9:13 PM Ed Sabol <[email protected]> wrote:
> On Apr 28, 2025, at 1:24 PM, Gambhir Singh <[email protected]>
> wrote:
> > Row Count - 50 Billion
>
> I've never dealt with a table that huge personally, but my concern would
> be that ALTER TABLE will lock the table for a very long time. Is this in a
> production environment with active usage of this table? Just SELECTs or are
> we talking UPDATEs and INSERTs as well? If so, you might need to do
> something more complicated than just ALTER TABLE.
>
> If you have enough disk space in the storage area for this database to
> have two identical copies of this 50 billion row table (with indexes!), you
> could make a copy of the table and either ALTER that copy or add the new
> column at the same time as making the copy and then, in a single
> transaction, rename the two tables to swap them. If you do it this way, the
> new table will replace the old table seamlessly without interrupting usage
> of the table. Somewhere in there, you might need to re-sync the two tables
> to make sure any rows that got inserted or updated while you were making
> the copy are incorporated into the new version of the table as well.
>
> Just some initial thoughts on how I would accomplish this and things I
> would consider when deciding how to do it.
>
>
COPY TO of that table, and then COPY FROM into a new table would let OP
experiment. Since it's 50Bn rows, COPY TO of a quarter of the rows is
probably adequate.
Hopefully this bolding comes through:
"When a column is added with *ADD COLUMN and a non-volatile DEFAULT* is
specified, the default is evaluated at the time of the statement and the
result stored in the table's metadata. That value will be used for the
column for all existing rows. If no DEFAULT is specified, NULL is used. *In
neither case is a rewrite of the table required.*"
According to https://www.postgresql.org/docs/17/sql-altertable.html, "*Adding
a *CHECK or *NOT NULL* constraint requires scanning the table to verify
that existing rows meet the constraint, but *does not require a table
rewrit*e."
That's probably pretty fast, even if an exclusive lock is required.
Thus, I'd probably try this on the table copy:
ALTER TABLE foo ADD COLUMN bar BIGINT NOT NULL DEFAULT 0;
UPDATE foo SET bar = 0 WHERE pk between 0*1000+0 AND 0*1000+9999;
UPDATE foo SET bar = 0 WHERE pk between 1*1000+0 AND 1*1000+9999;
UPDATE foo SET bar = 0 WHERE pk between 2*1000+0 AND 2*1000+9999;
etc.
The UPDATE statement would be in a bash loop, with the 0, 1, 2, 3... a
variable.
I'd also stick an occasional VACUUM in the bash script.
--
Death to <Redacted>, and butter sauce.
Don't boil me, I'm still alive.
<Redacted> lobster!
reply
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: [email protected]
Cc: [email protected]
Subject: Re: Adding New Column with default value.
In-Reply-To: <CANzqJaBdQ5CsvXzDNWd19-auoLyMT0V7UHjKvXd8Bp4T0oeNZg@mail.gmail.com>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox