[Resharding] Update sequences

[levkk]

No description provided.

[vitabaks]

@levkk maybe this will help you, this is how I usually copy all sequences:

# Copy sequences from source database
psql -h <source_dbhost> \
-U postgres \
-d <source_dbname>  \
-tAXc "
  select format(
    'select setval(%L, %s);',
    format('%I.%I', schemaname, sequencename),
    coalesce(last_value, 1) + 1000000
  )
  from pg_sequences
" > pg_sequences_increase.sql

# Apply new sequence values on target database
psql -U postgres -d <target_dbname> -f pg_sequences_increase.sql

Please note that here we copy the current value of all the sequences and add 1 million to them, usually this is enough, but in any case it is individual for each cluster because it depends on the load.

[vitabaks]

Although if you update a sequence only for a specific table, you won't need it.

[levkk]

I'm using the primary schema dump as the source of truth and separating statements into 3 phases:

• pre logical replication sync (tables, defaults, primary key index)
• post logical replication sync (foreign keys, secondary indexes)
• cutover sync (set sequence id to max(id) + 1)