[FIXED] Set all NULL `lock_version` values to 0 PART 2 (!30305) · Merge requests · GitLab.org / GitLab

Mario de la Ossa requested to merge 28560_cleanup_optimistic_locking_db_pt2-second-try into master Apr 23, 2020

What does this MR do?

This is part 2 of !18418 (merged)

This is a re-do from !25396 (merged) as there was a problem with a missing index

In a background migration, sets the value for lock_version to 0 if it's nil in the following tables:

ci_builds
ci_pipelines
ci_stages

Output of migration

Up:

== 20200217210353 CleanupOptimisticLockingNullsPt2: migrating =================
-- transaction_open?()
   -> 0.0000s
-- index_exists?(:ci_stages, :lock_version, {:where=>"lock_version IS NULL", :algorithm=>:concurrently})
   -> 0.0044s
-- execute("SET statement_timeout TO 0")
   -> 0.0003s
-- add_index(:ci_stages, :lock_version, {:where=>"lock_version IS NULL", :algorithm=>:concurrently})
   -> 0.0065s
-- execute("RESET ALL")
   -> 0.0002s
-- transaction_open?()
   -> 0.0000s
-- index_exists?(:ci_stages, :id, {:where=>"lock_version IS NULL", :name=>"tmp_index_ci_stages_lock_version", :algorithm=>:concurrently})
   -> 0.0029s
-- execute("SET statement_timeout TO 0")
   -> 0.0002s
-- add_index(:ci_stages, :id, {:where=>"lock_version IS NULL", :name=>"tmp_index_ci_stages_lock_version", :algorithm=>:concurrently})
   -> 0.0027s
-- execute("RESET ALL")
   -> 0.0002s
-- transaction_open?()
   -> 0.0000s
-- index_exists?(:ci_builds, :lock_version, {:where=>"lock_version IS NULL", :algorithm=>:concurrently})
   -> 0.0138s
-- execute("SET statement_timeout TO 0")
   -> 0.0003s
-- add_index(:ci_builds, :lock_version, {:where=>"lock_version IS NULL", :algorithm=>:concurrently})
   -> 0.0066s
-- execute("RESET ALL")
   -> 0.0002s
-- transaction_open?()
   -> 0.0000s
-- index_exists?(:ci_builds, :id, {:where=>"lock_version IS NULL", :name=>"tmp_index_ci_builds_lock_version", :algorithm=>:concurrently})
   -> 0.0131s
-- execute("SET statement_timeout TO 0")
   -> 0.0002s
-- add_index(:ci_builds, :id, {:where=>"lock_version IS NULL", :name=>"tmp_index_ci_builds_lock_version", :algorithm=>:concurrently})
   -> 0.0030s
-- execute("RESET ALL")
   -> 0.0002s
-- transaction_open?()
   -> 0.0000s
-- index_exists?(:ci_pipelines, :lock_version, {:where=>"lock_version IS NULL", :algorithm=>:concurrently})
   -> 0.0065s
-- execute("SET statement_timeout TO 0")
   -> 0.0002s
-- add_index(:ci_pipelines, :lock_version, {:where=>"lock_version IS NULL", :algorithm=>:concurrently})
   -> 0.0031s
-- execute("RESET ALL")
   -> 0.0003s
-- transaction_open?()
   -> 0.0000s
-- index_exists?(:ci_pipelines, :id, {:where=>"lock_version IS NULL", :name=>"tmp_index_ci_pipelines_lock_version", :algorithm=>:concurrently})
   -> 0.0074s
-- execute("SET statement_timeout TO 0")
   -> 0.0002s
-- add_index(:ci_pipelines, :id, {:where=>"lock_version IS NULL", :name=>"tmp_index_ci_pipelines_lock_version", :algorithm=>:concurrently})
   -> 0.0026s
-- execute("RESET ALL")
   -> 0.0002s
== 20200217210353 CleanupOptimisticLockingNullsPt2: migrated (0.3949s) ========

Down:

== 20200217210353 CleanupOptimisticLockingNullsPt2: reverting =================
-- transaction_open?()
   -> 0.0000s
-- index_exists?(:ci_stages, :lock_version, {:where=>"lock_version IS NULL", :algorithm=>:concurrently})
   -> 0.0062s
-- execute("SET statement_timeout TO 0")
   -> 0.0003s
-- remove_index(:ci_stages, {:where=>"lock_version IS NULL", :algorithm=>:concurrently, :column=>:lock_version})
   -> 0.0065s
-- execute("RESET ALL")
   -> 0.0002s
-- transaction_open?()
   -> 0.0000s
-- index_exists?(:ci_stages, :id, {:where=>"lock_version IS NULL", :name=>"tmp_index_ci_stages_lock_version", :algorithm=>:concurrently})
   -> 0.0036s
-- execute("SET statement_timeout TO 0")
   -> 0.0003s
-- remove_index(:ci_stages, {:where=>"lock_version IS NULL", :name=>"tmp_index_ci_stages_lock_version", :algorithm=>:concurrently, :column=>:id})
   -> 0.0047s
-- execute("RESET ALL")
   -> 0.0002s
-- transaction_open?()
   -> 0.0000s
-- index_exists?(:ci_builds, :lock_version, {:where=>"lock_version IS NULL", :algorithm=>:concurrently})
   -> 0.0144s
-- execute("SET statement_timeout TO 0")
   -> 0.0003s
-- remove_index(:ci_builds, {:where=>"lock_version IS NULL", :algorithm=>:concurrently, :column=>:lock_version})
   -> 0.0163s
-- execute("RESET ALL")
   -> 0.0002s
-- transaction_open?()
   -> 0.0000s
-- index_exists?(:ci_builds, :id, {:where=>"lock_version IS NULL", :name=>"tmp_index_ci_builds_lock_version", :algorithm=>:concurrently})
   -> 0.0135s
-- execute("SET statement_timeout TO 0")
   -> 0.0003s
-- remove_index(:ci_builds, {:where=>"lock_version IS NULL", :name=>"tmp_index_ci_builds_lock_version", :algorithm=>:concurrently, :column=>:id})
   -> 0.0162s
-- execute("RESET ALL")
   -> 0.0003s
-- transaction_open?()
   -> 0.0000s
-- index_exists?(:ci_pipelines, :lock_version, {:where=>"lock_version IS NULL", :algorithm=>:concurrently})
   -> 0.0101s
-- execute("SET statement_timeout TO 0")
   -> 0.0002s
-- remove_index(:ci_pipelines, {:where=>"lock_version IS NULL", :algorithm=>:concurrently, :column=>:lock_version})
   -> 0.0101s
-- execute("RESET ALL")
   -> 0.0004s
-- transaction_open?()
   -> 0.0000s
-- index_exists?(:ci_pipelines, :id, {:where=>"lock_version IS NULL", :name=>"tmp_index_ci_pipelines_lock_version", :algorithm=>:concurrently})
   -> 0.0093s
-- execute("SET statement_timeout TO 0")
   -> 0.0002s
-- remove_index(:ci_pipelines, {:where=>"lock_version IS NULL", :name=>"tmp_index_ci_pipelines_lock_version", :algorithm=>:concurrently, :column=>:id})
   -> 0.0092s
-- execute("RESET ALL")
   -> 0.0002s
== 20200217210353 CleanupOptimisticLockingNullsPt2: reverted (0.1249s) ========

Time calculations

For each of these I performed EXPLAIN SELECT count(*) table_name WHERE lock_version IS NULL

Batches are 10,000 rows each

CI Builds table

Aggregate  (cost=213383199.17..213383199.18 rows=1 width=8)
  ->  Seq Scan on ci_builds  (cost=0.00..213366227.28 rows=6788756 width=0)
        Filter: (lock_version IS NULL)

6788756/10000 = ~679 batches ~1358 minutes = ~23 hours

CI Pipelines

Aggregate  (cost=12590176.44..12590176.45 rows=1 width=8)
  ->  Seq Scan on ci_pipelines  (cost=0.00..12576426.08 rows=5500145 width=0)
        Filter: (lock_version IS NULL)

5500145/10000 = 551 batches ~1,102 minutes = ~19 hours

CI Stages

Aggregate  (cost=10342082.50..10342082.51 rows=1 width=8)
  ->  Seq Scan on ci_stages  (cost=0.00..10308197.88 rows=13553849 width=0)
        Filter: (lock_version IS NULL)

13553849/10000 = 1,356 batches ~2712 minutes = ~45 hours

Screenshots

Does this MR meet the acceptance criteria?

Conformity

Availability and Testing

Review and add/update tests for this feature/bug. Consider all test levels. See the Test Planning Process.
Tested in all supported browsers
Informed Infrastructure department of a default or new setting change, if applicable per definition of done

Security

If this MR contains changes to processing or storing of credentials or tokens, authorization and authentication methods and other items described in the security review guidelines:

Label as security and @ mention @gitlab-com/gl-security/appsec
The MR includes necessary changes to maintain consistency between UI, API, email, or other methods
Security reports checked/validated by a reviewer from the AppSec team

[FIXED] Set all NULL `lock_version` values to 0 PART 2