insert into t (select * from t) bogus in multithreaded execution #2883
Milestone
Comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
{{ message }}
Date: 2011-09-15 15:45:55 +0200
From: @sjoerdmullender
To: MonetDB5 devs <>
Version: 11.5.1 (Aug2011) [obsolete]
CC: @mlkersten, @yzchang
Blocker for: #2882
Last updated: 2012-11-27 14:44:10 +0100
Comment 16225
Date: 2011-09-15 15:45:55 +0200
From: @sjoerdmullender
Created attachment 73
trace of last insert query
On a multicore system, dataflow can cause that the subject query inserts more into the table than it should.
Using mserver5 --forcemito do:
create table t (i int);
insert into t values (1);
insert into t (select * from t);
insert into t (select * from t);
After these queries there should be 4 elements in t, but I get 6. The reason is that the insertions table is appended to first (assignment to X_44 in trace) before that same table is used to find old inserted values that have to be copied (assignment to X_9 in trace).
Comment 16227
Date: 2011-09-15 23:39:42 +0200
From: @mlkersten
The cause are the multiple mitosis plans, who each
insert the tuples. All updates should be collected
in a temporary BAT before being inserted into the persistent version.
Comment 16228
Date: 2011-09-16 10:13:42 +0200
From: @sjoerdmullender
Changeset 7e2a2426e682 made by Sjoerd Mullender sjoerd@acm.org in the MonetDB repo, refers to this bug.
For complete details, see http//devmonetdborg/hg/MonetDB?cmd=changeset;node=7e2a2426e682
Changeset description:
Comment 16286
Date: 2011-09-16 15:48:57 +0200
From: @sjoerdmullender
Changeset c22ffd42d642 made by Sjoerd Mullender sjoerd@acm.org in the MonetDB repo, refers to this bug.
For complete details, see http//devmonetdborg/hg/MonetDB?cmd=changeset;node=c22ffd42d642
Changeset description:
Comment 16287
Date: 2011-09-16 15:49:50 +0200
From: @sjoerdmullender
The bug has been fixed. No separate test is needed since the test for bug #2882 covers this issue as well.
Comment 16297
Date: 2011-09-19 15:00:18 +0200
From: @grobian
Could this fix be the cause of orphans on intel Darwin, and complete defunct processes on Solaris? (I just manually killed Python/Mtest, it doesn't seem to reap its childeren correctly)
Comment 16350
Date: 2011-09-30 10:58:44 +0200
From: @grobian
Released in Aug2011-SP1
Comment 18090
Date: 2012-11-27 14:44:10 +0100
From: @yzchang
Covered by the test for bug #2882:
sql/test/BugTracker-2011/Tests/delete-large-table.Bug-2882.sql
The text was updated successfully, but these errors were encountered: