external
/
citus
mirror of https://github.com/citusdata/citus.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
3,180 Commits
636 Branches
162 Tags
99 MiB
Commit Graph

1989 Commits (fd17e4578e2238feb32e5fa3e54c242560e172a5)

Author SHA1 Message Date
Onder Kalaci fd17e4578e Improve tests 2020-01-17 16:02:57 +01:00
Onder Kalaci 0bf1e81e33 Cache local plans on BeginScan 2020-01-17 16:02:57 +01:00
Onder Kalaci 08d148d43e Make TaskAccessesLocalNode external function 2020-01-17 16:02:57 +01:00
Onder Kalaci 5dc454cdad Exclude localPlannedStatements from copy distributedPlan 2020-01-17 16:02:57 +01:00
Onder Kalaci ff12df411b Add LocalPlannedStatement struct 2020-01-17 16:02:57 +01:00
Onder Kalaci 016f561e45 Ingest data for cte_inline tests 2020-01-17 12:46:00 +01:00
Onder Kalaci 3833a7e686 Fix issues for CTE inlining on Postgres 11 
Comment from code:

/*
 * We had to implement this hack because on Postgres11 and below, the originalQuery
 * and the query would have significant differences in terms of CTEs where CTEs
 * would not be inlined on the query (as standard_planner() wouldn't inline CTEs
 * on PG 11 and below).
 *
 * Instead, we prefer to pass the inlined query to the distributed planning. We rely
 * on the fact that the query includes subqueries, and it'd definitely go through
 * query pushdown planning. During query pushdown planning, the only relevant query
 * tree is the original query.
 */
 2020-01-17 11:59:02 +01:00
Jelte Fennema 246435be7e
Lazy query deparsing executable queries (#3350) 
Deparsing and parsing a query can be heavy on CPU. When locally executing 
the query we don't need to do this in theory most of the time.

This PR is the first step in allowing to skip deparsing and parsing
the query in these cases, by lazily creating the query string and
storing the query in the task. Future commits will make use of this and
not deparse and parse the query anymore, but use the one from the task
directly.
 2020-01-17 11:49:43 +01:00
Hadi Moshayedi 6cf1c01660 Don't use repartitioned INSERT/SELECT for repartition joins 2020-01-16 23:40:31 -08:00
Hadi Moshayedi 5eeb07124f Repartitioned INSERT/SELECT: include job id in result id prefix 2020-01-16 23:24:52 -08:00
Hadi Moshayedi a079278b0c Repartitioned INSERT/SELECT: Add a GUC to enable/disable it 2020-01-16 23:24:52 -08:00
Hadi Moshayedi ce5eea4885 INSERT/SELECT: make SELECT column names unique 2020-01-16 23:24:52 -08:00
Hadi Moshayedi 3258d87f3e Isolation tests for INSERT/SELECT repartition 2020-01-16 23:24:52 -08:00
Hadi Moshayedi 8b27a9a195 More range partitioned tests 2020-01-16 23:24:52 -08:00
Hadi Moshayedi 8635396cea Repartitioned INSERT/SELECT: Test rollback behaviour 2020-01-16 23:24:52 -08:00
Hadi Moshayedi 43218eebf6 Failure tests for INSERT/SELECT repartition 2020-01-16 23:24:52 -08:00
Hadi Moshayedi 665b33dca1 MX tests for INSERT/SELECT repartition 2020-01-16 23:24:52 -08:00
Hadi Moshayedi af2349f21f Repartitioned INSERT/SELECT: Add a prepared statement test 2020-01-16 23:24:52 -08:00
Hadi Moshayedi 97072c9eb1 INSERT/SELECT: show method in EXPLAIN output 2020-01-16 23:24:52 -08:00
Hadi Moshayedi b143d9588a Repartitioned INSERT/SELECT: Test GROUP BY 2020-01-16 23:24:52 -08:00
Hadi Moshayedi fe548b762f Repartitioned INSERT/SELECT: Test CTEs 2020-01-16 23:24:52 -08:00
Hadi Moshayedi 494cc383cc Repartitioned INSERT/SELECT: Enable RETURNING 2020-01-16 23:24:52 -08:00
Hadi Moshayedi 4b14347fc3 Tests for DML followed by insert/select repartition 2020-01-16 23:24:52 -08:00
Hadi Moshayedi 44a2aede16 Don't start a coordinated transaction on workers. 
Otherwise transaction hooks of Citus kick in and might cause unwanted errors.
 2020-01-16 23:24:52 -08:00
Hadi Moshayedi 42c3c03b85 Handle extra columns added in ExpandWorkerTargetEntry() in repartitioned INSERT/SELECT 2020-01-16 23:24:52 -08:00
Hadi Moshayedi 89463f9760 Repartitioned INSERT/SELECT: cast columns in SELECT targets 2020-01-16 23:24:52 -08:00
Hadi Moshayedi d67a384350 Enable repartitioned INSERT/SELECT ON CONFLICT. 2020-01-16 23:24:52 -08:00
Hadi Moshayedi b4e5f4b10a Implement INSERT ... SELECT with repartitioning 2020-01-16 23:24:52 -08:00
Hadi Moshayedi ced876358d INSERT/SELECT: Refactor out AddInsertSelectCasts 2020-01-16 23:24:52 -08:00
Hadi Moshayedi d449c1857c INSERT/SELECT: Use ExecutePlan* instead of ExecuteSelect* 2020-01-16 23:24:52 -08:00
Hadi Moshayedi e30580e2bd Add ORDER BY to multi_row_insert.sql 2020-01-16 15:20:39 -08:00
Jelte Fennema 0ee1eab070 Make tests fail with a useful error message 2020-01-16 18:30:30 +01:00
Jelte Fennema cb5154cf03 Add more failing tests, of which some have bad error messages 2020-01-16 18:30:30 +01:00
Marco Slot 82f1fffa28 Fix epoll_ctl() error message on connection error 2020-01-16 06:40:57 +01:00
Onder Kalaci dc17c2658e Defer shard pruning for fast-path router queries to execution 
This is purely to enable better performance with prepared statements.
Before this commit, the fast path queries with prepared statements
where the distribution key includes a parameter always went through
distributed planning. After this change, we only go through distributed
planning on the first 5 executions.
 2020-01-16 16:59:36 +01:00
Onder Kalaci 933d666c0d Do not forget to copy fastPathRouterPlan@DistributedPlan 2020-01-16 16:39:20 +01:00
Halil Ozan Akgul c5539d20d9 Adds alter table schema propagation 2020-01-16 17:04:16 +03:00
Nils Dijk b6e09eb691
Fix: distributed function with table reference in declare (#3384) 
DESCRIPTION: Fixes a problem when adding a new node due to tables referenced in a functions body

Fixes #3378 

It was reported that `master_add_node` would fail if a distributed function has a table name referenced in its declare section of the body. By default postgres validates the body of a function on creation. This is not a problem in the normal case as tables are replicated to the workers when we distribute functions.

However when a new node is added we first create dependencies on the workers before we try to create any tables, and the original tables get created out of bound when the metadata gets synced to the new node. This causes the function body validator to raise an error the table is not on the worker.

To mitigate this issue we set `check_function_bodies` to `off` right before we are creating the function.

The added test shows this does resolve the issue. (issue can be reproduced on the commit without the fix)
 2020-01-16 14:21:54 +01:00
Jelte Fennema e76281500c
Replace shardId lock with lock on colocation+shardIntervalIndex (#3374) 
This new locking pattern makes sure that some deadlocks that could
happend during rebalancing cannot occur anymore.
 2020-01-16 13:14:01 +01:00
Jelte Fennema 86343bcc8f Re-add test that broke with GUC workaround 2020-01-16 12:34:50 +01:00
Jelte Fennema 6b9b633695 Add more tests for prepared statements 2020-01-16 12:28:15 +01:00
Jelte Fennema 43a3fdd12f Fix comment 2020-01-16 12:28:15 +01:00
Jelte Fennema fe3827e499 Add tests for [NOT] MATERIALEZED 2020-01-16 12:28:15 +01:00
Onder Kalaci 326dfab44a Fix a query which triggers an existing bug, see https://github.com/citusdata/citus/issues/3189#issuecomment-571497051 2020-01-16 12:28:15 +01:00
Onder Kalaci 81d8178625 Note that we'll drop the GUC after PG 11 support dropped 2020-01-16 12:28:15 +01:00
Onder Kalaci c653923960 Update regression tests 6 
Local execution and CTE pushdown
 2020-01-16 12:28:15 +01:00
Onder Kalaci 3818be45a6 Update regression tests-5 
Failure tests that rely on intermediate results
 2020-01-16 12:28:15 +01:00
Onder Kalaci 1e85938b46 Update regression tests-4 
Update the MX tests. Similar to the previous commits, prevent CTE
inlining in some cases to prevent divergent test outputs.
 2020-01-16 12:28:15 +01:00
Onder Kalaci fc07bd7c5b Update regression tests-3 
Update the regression tests which only change in PG 12.
 2020-01-16 12:28:15 +01:00
Onder Kalaci 64560b07be Update regression tests-2 
In this commit, we're introducing a way to prevent CTE inlining via a GUC.

The GUC is used in all the tests where PG 11 and PG 12 tests would diverge
otherwise.

Note that, in PG 12, the restriction information for CTEs are generated. It
means that for some queries involving CTEs, Citus planner (router planner/
pushdown planner) may behave differently. So, via the GUC, we prevent
tests to diverge on PG 11 vs PG 12.

When we drop PG 11 support, we should get rid of the GUC, and mark
relevant ctes as MATERIALIZED, which does the same thing.
 2020-01-16 12:28:15 +01:00