> Yeah, I thought the int --> bigint would not do a table rewrite. Testing
> showed otherwise. Forget that idea.
Got it. Not sure what else we should consider. It seemed like the constraint might be possible, but currently need a far bigger table to be able to tell for sure, since we can't explain a DDL.
On 7/21/20 2:18 PM, Mohamed Wael Khobalatte wrote:
>> > test_(aklaver)5432> alter table change_seq alter COLUMN id set data type
>> bigint;
>> ALTER TABLE
>> test_(aklaver)5432> \d change_seq
>> Table "public.change_seq"
>> Column | Type | Collation | Nullable | Default
>> --------+--------+-----------+----------+----------------------------------------
>>
>> id | bigint | | not null |
>> nextval('change_seq_id_seq'::regclass)
>> Indexes:
>> "change_seq_pkey" PRIMARY KEY, btree (id)
>
> This is significant downtime, since it locks exclusively, no? We want to
> avoid that.
Yeah, I thought the int --> bigint would not do a table rewrite. Testing
showed otherwise. Forget that idea.
>
> > Side note- EOL for 9.6 is coming next year so just a plug for
> upgrading when possible, perhaps utilizing pglogical to get to v11 or 12.
>
> Yep, we are painfully aware. The id growth will beat us to it, so we
> need to deal with that first.
>
>
--
Adrian Klaver
adrian.klaver@aklaver.com