ed670156e4
Looking at our hot queries (i.e. frequency * duration = long), it
appears we were seq scanning and disqualifying nearly all `semaphore`
records running a query like:
SELECT * FROM "semaphore" WHERE ("strand_id" = $1)
The result was excessive consumed CPU time.
The use of `CREATE INDEX CONCURRENTLY` is excessive, but it makes one
of our inherited RuboCop rules happy. Maybe I'll disable that rule
later.
12 lines
177 B
Ruby
12 lines
177 B
Ruby
# frozen_string_literal: true
|
|
|
|
Sequel.migration do
|
|
no_transaction
|
|
|
|
change do
|
|
alter_table(:semaphore) do
|
|
add_index :strand_id, concurrently: true
|
|
end
|
|
end
|
|
end
|