postgres-xc-committers Mailing List for Postgres-XC (Page 2)

Project "Postgres-XC".

The branch, master has been updated
       via  0f6b92a8aaee2a33e416ebcb2057a5f97aa8b282 (commit)
      from  f3f0a8bc7f4e4c76dc4974cee1397404e14ca6fd (commit)


- Log -----------------------------------------------------------------
http://postgres-xc.git.sourceforge.net/git/gitweb.cgi?p=postgres-xc/postgres-xc;a=commitdiff;h=0f6b92a8aaee2a33e416ebcb2057a5f97aa8b282

commit 0f6b92a8aaee2a33e416ebcb2057a5f97aa8b282
Author: Michael P <mic...@us...>
Date:   Tue Oct 11 08:20:04 2011 +0900

    Stabilize serial regressions with multiple Coordinators
    
    This resulted in a snapshot issue that may crash a remote Coordinator
    due to assertion failure on relations cached.
    It was only reproducible with multiple Coordinators, and still this
    assertion error appeared randomly.

M	src/backend/storage/ipc/procarray.c
M	src/test/regress/parallel_schedule

-----------------------------------------------------------------------

Summary of changes:
 src/backend/storage/ipc/procarray.c |    4 ++--
 src/test/regress/parallel_schedule  |    5 ++++-
 2 files changed, 6 insertions(+), 3 deletions(-)


hooks/post-receive
-- 
Postgres-XC

Project "Postgres-XC".

The branch, master has been updated
       via  f3f0a8bc7f4e4c76dc4974cee1397404e14ca6fd (commit)
      from  34ba88b93acf5dd8f06f8dd1b13061d4ca498ab2 (commit)


- Log -----------------------------------------------------------------
http://postgres-xc.git.sourceforge.net/git/gitweb.cgi?p=postgres-xc/postgres-xc;a=commitdiff;h=f3f0a8bc7f4e4c76dc4974cee1397404e14ca6fd

commit f3f0a8bc7f4e4c76dc4974cee1397404e14ca6fd
Author: Pavan Deolasee <pav...@gm...>
Date:   Wed Oct 5 16:35:25 2011 +0530

    Fix a bug in compute_hash(). We should use DatumGetInt64() for INT8OID
    types while computing hash value. This was leading to undeterministic
    behavior while dealing with int8 distribution columns in a table.

M	src/backend/access/hash/hashfunc.c

-----------------------------------------------------------------------

Summary of changes:
 src/backend/access/hash/hashfunc.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)


hooks/post-receive
-- 
Postgres-XC

Project "Postgres-XC".

The branch, master has been updated
       via  34ba88b93acf5dd8f06f8dd1b13061d4ca498ab2 (commit)
      from  9875b6a3170e02666ad2e4ad12460dd412800598 (commit)


- Log -----------------------------------------------------------------
http://postgres-xc.git.sourceforge.net/git/gitweb.cgi?p=postgres-xc/postgres-xc;a=commitdiff;h=34ba88b93acf5dd8f06f8dd1b13061d4ca498ab2

commit 34ba88b93acf5dd8f06f8dd1b13061d4ca498ab2
Author: Pavan Deolasee <pav...@gm...>
Date:   Tue Oct 4 13:49:44 2011 +0530

    Fix a server crash is query deparsing. PostgreSQL has introduced a new
    API for query deparse since 9.1, but that works only for init-ed plans.
    We were using that API on a Plan node where as it expects a PlanState node
    to work on. This causes segfault, especially on a 32-bit machine, but the
    code was wrong anyways.
    
    We create dummy PlanState nodes for the purpose of deparse. Since the
    use is quite limited right now, it works. But this is just a work-around
    and in the long term, we should either add appropriate API for query
    deparsing at planning time or move the logic of generating query string at
    the execution time. The second option looks more promising, but the remote
    join reduction logic may need some rework.

M	src/backend/optimizer/plan/createplan.c
M	src/backend/pgxc/plan/planner.c
M	src/backend/pgxc/pool/postgresql_fdw.c
M	src/backend/utils/adt/ruleutils.c
M	src/include/utils/builtins.h

-----------------------------------------------------------------------

Summary of changes:
 src/backend/optimizer/plan/createplan.c |    2 +-
 src/backend/pgxc/plan/planner.c         |    2 +-
 src/backend/pgxc/pool/postgresql_fdw.c  |    2 +-
 src/backend/utils/adt/ruleutils.c       |  104 +++++++++++++++++++++++++++++++
 src/include/utils/builtins.h            |    4 +
 5 files changed, 111 insertions(+), 3 deletions(-)


hooks/post-receive
-- 
Postgres-XC

Project "Postgres-XC".

The branch, master has been updated
       via  9875b6a3170e02666ad2e4ad12460dd412800598 (commit)
      from  a3ddbdb7739fd2f6deb5c196b0d42bee23f01664 (commit)


- Log -----------------------------------------------------------------
http://postgres-xc.git.sourceforge.net/git/gitweb.cgi?p=postgres-xc/postgres-xc;a=commitdiff;h=9875b6a3170e02666ad2e4ad12460dd412800598

commit 9875b6a3170e02666ad2e4ad12460dd412800598
Author: Michael P <mic...@us...>
Date:   Tue Oct 4 08:28:00 2011 +0900

    Fix for an assertion failure when running regressions
    
    The origin of the assertion error was double:
    1) DISCARD queries send directly from the pooler did not include
    transaction IDs necessary also on remote nodes. This finishes
    by having a transaction committing with a local transaction ID,
    not a global one.
    2) When a connection is cut between 2 Postgres-XC nodes and session
    on remote node hold temporary objects, a transaction ID was consumed
    to clean up the temporary objects with a callback API. This transaction
    finished using a local transaction ID, where it should have used a global
    one.
    
    The use of local transaction IDs for commit does not protect commit
    consistency at all and may create issues with transaction IDs committed
    twice in the worse case. Sometimes it also created issues with
    assertions in 2PC commit.
    
    Some simplifications in pooler are added, and to insure that temporary
    objects are correctly deleted from the system at each session ending,
    connections holding temporary objects are not pull back to pull but simply
    destroyed.
    There may be some more performant way to do that, but it may involve to send
    on all the nodes incolved in session a "DISCARD TEMP" query at least with a
    transaction ID with a snapshot, and I am not sure this is worth doing at
    session ending on Coordinator. So for the time being only cut connection.
    Connections that used only session parameters or normal objects except
    temporary ones are put back normally to pooler.

M	src/backend/catalog/namespace.c
M	src/backend/pgxc/pool/execRemote.c
M	src/backend/pgxc/pool/pgxcnode.c
M	src/backend/pgxc/pool/poolmgr.c
M	src/backend/tcop/utility.c
M	src/include/pgxc/poolmgr.h

-----------------------------------------------------------------------

Summary of changes:
 src/backend/catalog/namespace.c    |   16 +++
 src/backend/pgxc/pool/execRemote.c |    2 +-
 src/backend/pgxc/pool/pgxcnode.c   |   38 ++----
 src/backend/pgxc/pool/poolmgr.c    |  231 +++++++++---------------------------
 src/backend/tcop/utility.c         |   18 +--
 src/include/pgxc/poolmgr.h         |    2 +-
 6 files changed, 95 insertions(+), 212 deletions(-)


hooks/post-receive
-- 
Postgres-XC

Project "Postgres-XC".

The branch, master has been updated
       via  a3ddbdb7739fd2f6deb5c196b0d42bee23f01664 (commit)
      from  dfa5b5171b09930bd6f2863bfb13cfd91c000d1f (commit)


- Log -----------------------------------------------------------------
http://postgres-xc.git.sourceforge.net/git/gitweb.cgi?p=postgres-xc/postgres-xc;a=commitdiff;h=a3ddbdb7739fd2f6deb5c196b0d42bee23f01664

commit a3ddbdb7739fd2f6deb5c196b0d42bee23f01664
Author: Pavan Deolasee <pav...@gm...>
Date:   Mon Oct 3 14:56:05 2011 +0530

    We were wrongly type-casting many other plan nodes as RemoteQuery,
    even though there were not. This was causing segfaults, especially on
    32-bit machines. Fixed by this patch.

M	src/backend/commands/explain.c

-----------------------------------------------------------------------

Summary of changes:
 src/backend/commands/explain.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)


hooks/post-receive
-- 
Postgres-XC