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@xxxxxxxxxxx
