Commit Graph

953 Commits (581c8c02cc1b4fe5149276f1bb7be39c4d1647f9)

Author SHA1 Message Date
Marco Slot 3a4d5f8182 Remove filter checks on leaf queries 2017-11-30 12:25:14 +01:00
Marco Slot 3f03cb6a6a Support UNION with joins in the subqueries 2017-11-30 10:37:56 +01:00
Marco Slot a9933deac6 Make real time executor work in transactions 2017-11-30 09:59:32 +03:00
mehmet furkan şahin 6041f85b70
Add tests for non-propagated VACUUM/ANALYZE 2017-11-29 16:06:50 -07:00
Jason Petersen 0eacf6bd95
Refactor VacuumStmt checker to be single-return
Decided this would be safer for the future (defaults to unsupported).
2017-11-29 16:06:50 -07:00
Jason Petersen b12e77ab0e
Ensure unsupported VACUUMs don't go to workers
Apparently these two blocks have been incorrect for nearly a year…
2017-11-29 16:06:50 -07:00
Marco Slot 7ea718fd8d Round-robin over worker nodes for 0-shard router queries 2017-11-29 15:52:22 +01:00
Marco Slot ae67fa0e52 Do not run multi_mx_modifications in parallel with multi_mx_transaction_recovery 2017-11-29 15:35:21 +01:00
mehmet furkan şahin b6eb0c2823 multi_subquery_behavioral_analytics.sql query fix by adding proper order by 2017-11-28 14:15:46 +03:00
mehmet furkan şahin 1b06b2b306 The data used in regression tests is reduced
This commit reduces the size of the data in users_table.data
and events_table.data from 10K rows to 100 rows.
2017-11-28 14:15:46 +03:00
Onder Kalaci 05fb0dd020 Add infrastructure for filtering restriction contexts based on the input query
In subquery pushdown, we first ensure that each relation is joined with at least
on another relation on the partition keys. That's fine given that the decision
is binary: pushdown the query at all or not.

With recursive planning, we'd want to check whether any specific part
of the query can be pushded down or not. Thus, we need the ability to
understand which part(s) of the subquery is safe to pushdown. This commit
adds the infrastructure for doing that.
2017-11-28 09:58:21 +02:00
Onder Kalaci 26d9b58e9e Make sure that ExtractRangeTableRelationWalker never misses RTE_RELATION 2017-11-28 09:27:34 +02:00
Onder Kalaci 32def06ebd Split assigning RTE identities and partitioning related query modifications
Note that we used to iterate over the RTEs once for performance reasons.
However, keeping an extra copy of original query seems more costly and
hard to maintain/explain.
2017-11-28 09:27:34 +02:00
Marco Slot feffe86440 Subqueries containing functions go through subquery pushdown 2017-11-27 22:13:02 +01:00
Onder Kalaci 48f96bf3e5 Enable non equi joins in subquery pushdown
Subquery pushdown planning is based on relation restriction
equivalnce. This brings us the opportuneatly to allow any
other joins as long as there is an already equi join between
the distributed tables.

We already allow that for joins with reference tables and
this commit allows that for joins among distributed tables.
2017-11-23 16:13:46 +02:00
mehmet furkan şahin 032b34ea52 some more parallelization 2017-11-23 14:10:42 +03:00
Onder Kalaci 16421f089f Register citus custom scan nodes 2017-11-23 11:38:33 +02:00
Onder Kalaci 83c1143505 Refactor custom scan related codes
In this commit, we don't change any codes, only create a new
file and move the related functions and types there.
2017-11-23 11:38:12 +02:00
Marco Slot 20a526d5c4 Fix memory leak in ListToHashSet 2017-11-22 11:26:58 +01:00
Marco Slot f4ceea5a3d Enable 2PC by default 2017-11-22 11:26:58 +01:00
Marco Slot 8486f76e15 Auto-recover 2PC transactions 2017-11-22 11:26:58 +01:00
Marco Slot 6ba3f42d23 Rename MultiPlan to DistributedPlan 2017-11-22 09:36:24 +01:00
Marco Slot 0ad39b36fe Treat immutable table functions and constant subqueries as reference tables 2017-11-21 14:15:22 +01:00
Onder Kalaci d558ebb923 Relax the checks on ensuring distribution columns for target entries
With this commit, we allow pushing down subqueries with only
reference tables where GROUP BY or DISTINCT clause or Window
functions include only columns from reference tables.
2017-11-21 12:28:14 +02:00
Andres Freund d063658d6d Protect some initializations from being called during backend startup.
On EXEC_BACKEND builds these functions shouldn't be called at every
backend start.
2017-11-20 15:29:51 -08:00
Brian Cloutier d267e0f9fa EXEC_BACKEND: don't put pointers to shared hashes into shared memory
Store pointers to shared hashes in process-local variables. Previously
pointers to shared hashes were put into shared memory. This causes
problems on EXEC_BACKEND because everybody calls execve and receives a
brand new address space; the shared hash will be in a different place
for every backend. (normally we call fork, which gives you a copy of the
address space, so these pointers remain constant)
2017-11-20 15:29:51 -08:00
Brian Cloutier 30a2365d81 Rename CreateDirectory to CitusCreateDirectory 2017-11-20 14:38:26 -08:00
Brian Cloutier aa2ab023a2 Rename RemoveDirectory -> CitusRemoveDirectory 2017-11-20 14:21:52 -08:00
Brian Cloutier 06f756b0a1 Rename DeleteFile -> CitusDeleteFile 2017-11-20 13:30:11 -08:00
mehmet furkan şahin 34709c2a16 Regression tests parallelization PART-1 2017-11-20 18:03:37 +03:00
Marco Slot 9793218122 Do not commit already-committed prepared transactions in recovery 2017-11-20 13:18:48 +01:00
Marco Slot fe798cf0f9 Add recovery vs. recovery isolation test 2017-11-20 12:26:25 +01:00
Marco Slot ae47df01ea Observe prepared xacts twice in RecoverWorkerTransactions to avoid race condition 2017-11-20 11:44:08 +01:00
Marco Slot 2410c2e450 Rewrite recover_prepared_transactions to be fast, non-blocking 2017-11-20 11:27:40 +01:00
mehmet furkan şahin 314fc09d90 regression test shard_count is changed from 32 to 4 2017-11-20 12:47:49 +03:00
mehmet furkan şahin 8d55754b4d the tests are separated and some more added 2017-11-20 11:45:48 +03:00
mehmet furkan şahin 636faadc47 create_distributed_table vs create_distributed_table, master_append_table_to_shard vs master_apply_delete_command, master_apply_delete_command vs master_apply_delete_command are added 2017-11-20 11:45:48 +03:00
mehmet furkan şahin 0722334e50 concurrent master_append_table test is added 2017-11-20 11:45:48 +03:00
mehmet furkan şahin f45988962f multi-shard update affecting the same/different rows 2017-11-20 11:45:48 +03:00
Onder Kalaci 5bea95009b Skip autovacuum processes for distributed deadlock detection
Autovacuum process cancels itself if any modification starts
on the table in order to avoid blocking your regular Postgres
sessions. That's normal and expected. Thus, any locks held by
autovacuum process cannot involve in a distributed deadlock
since it'll be released if needed.
2017-11-15 14:32:16 +02:00
Onder Kalaci c65c153a46 Skip speculative locks for distributed deadlock detection
These locks are held for a very short duration time and cannot
contribute to a deadlock. Speculative locks are used by Postgres
for internal notification mechanism among transactions.
2017-11-15 12:43:45 +02:00
Marco Slot bbbadd6d1b Bump Citus version to 7.2devel 2017-11-15 10:32:49 +01:00
Marco Slot ea306c6cfe Use citus.next_placement_id where practical in regression tests 2017-11-15 10:12:06 +01:00
Marco Slot d3b634b301 Allow generating placement IDs without using the sequence 2017-11-15 10:12:06 +01:00
Marco Slot 89eb833375 Use citus.next_shard_id where practical in regression tests 2017-11-15 10:12:05 +01:00
Marco Slot c24a0875a5 Allow generating shard IDs without using the sequence 2017-11-15 10:12:05 +01:00
Brian Cloutier 0f3230170f Pull in INT32_MAXINT and INT32_MININT 2017-11-14 14:03:46 -08:00
Brian Cloutier 0db8277266 remove unused errno import 2017-11-14 13:09:34 -08:00
Brian Cloutier 5d9f3ae7fd Remove unused poll import from multi_real_time_executor 2017-11-14 13:09:34 -08:00
Marco Slot 533a533565 Only drop sequences on workers with metadata 2017-11-14 16:01:56 +01:00