E.4. Release 1.0.1

E.4.1. Migration to Version 1.0.1

A dump/restore is not required for those running 1.0.X.

E.4.2. Changes

• Make data ordering consistent in query for test join.sql (Michael Paquier)

  Regression test join failed occasionally depending on the cluster configuration.

• Fix for sequence removal on GTM when related database is dropped (Michael Paquier, Dimitrije Radojevic)

  There was an error on GTM side when comparing sequence key strings related to key size when sequence drop is based on a sequence key type GTM_SEQ_DB_NAME.

• Fix for INSERT SELECT for partially replicated tables. Remote DML planning was not taking into account the node list to be used on a relation when modifying data on it. (Michael Paquier)

• Push -l option of gtm_ctl to gtm (Michael Meskes).

  This avoids to have an empty output file when gtm_ctl has a log file defined and change log file name of GTM according to what is given by gtm_ctl

• Fix for crash at gtm_proxy startup in 32b-linux machines (Koichi Suzuki, Plexo Rama)

  This fix adds an additional boolean flag in GTM-Proxy thread info to make it similar to the one used in GTM. The problem which was happening was that GTM-Proxy tried to use the memory structure of the thread info of GTM. Crash problems are solved but we should find in the future a smarter way to manage GTM and GTM-Proxy memory allocation APIs as they are now strongly dependant on each other and we may find similar issues if some new developments are done on GTM and/or GTM-Proxy.

• Activate gtm_ctl status (Koichi Suzuki, Michael Paquier)

  gtm_ctl can control the status of a GTM server like pg_ctl does with PostgreSQL instances (in the case of Postgres-XC Coordinator and Datanode).

• Fix for incorrect COPY for partially distributed tables (Michael Paquier)

  When a table was distributed on a portion of nodes. COPY process was not correctly choosing the output node during a COPY TO. This problem has been put into light by a case where table was replicated without using the first datanode of the cluster. It happened that there was the same problem for hash/modulo tables as well. A test case in xc_copy has been added to cover this problem, and a small portion managing COPY TO in execRemote.c has been cleaned.

• Correct whitespaces, format and message format in gtm_client.c (Andrei Martsinchyk)

  This had consequences on the data types of the values sent by GTM which was not matching those GTM Standby was trying to read.

• Solve SEGENV for GTM-Standby (Andrei Martsinchyk)

  Wrong variable was referenced when writing terminating null at the end of a string. It caused seg faults, because the referenced variable was not initialized yet.

• Introduce new GTM thread status managing thread backup (Andrei Martsinchyk)

  In the case of a GTM standby crash during the initial backup, the master GTM remained locked. The origin of this problem was during GTM backup process, where GTM locks all its threads one by one, performs a backup and then releases everything. This commit adds an additional control layer able to detect the thread status during backup. The new status GTM_THREAD_BACKUP is introduced for this purpose. Once backup is performed, the thread status is changed to GTM_THREAD_RUNNING. Hence, if the handler (GTM) is invoked upon sudden disconnection for backend (here GTM-Standby) disconnect of its connection and it sees that the current thread has status GTM_THREAD_BACKUP it goes ahead and releases the locks.

• Release GTM thread lock before exit (Andrei Martsinchyk)

  This avoid to have another thread waiting for it, as it is possible it waits for it.

• Block GXID generation on standbys (Andrei Martsinchyk)

  If a node accidentally connects to the Standby and reconnects to the Master afterwards, it will get an inconsistent GXID value since GTM Master is not aware of GXIDs generated by GTM Standby.

• validate_part_col_updatable was falsely assuming that the table with no location info is as good as non-existent table. This assumption is not true for catalogs or coordinator only tables. In such cases, this function should not look for any distribution column, since one can not exist for a table local to the coordinator (Ashutosh Bapat)

• Fix for bug 3556031: pgadmin3 connection error (Michael Paquier)

  pgadmin3 parses the output of SELECT version() to determine the server version it is connecting to. For Postgres-XC, this output result was changed and completed with Postgres-XC and PostgreSQL version. pgadmin thought that the version of server was the one of Postgres-XC, meaning 1.0 for 1.0 stable and 1.1 for master branch. In consequence to that, SQL queries with very old catalog were run to server, and of course errors were reported to client and pgadmin could not work correctly. This is solved with changing back version() to its former output. An additional function pgxc_version() is added to provide a way for application to inform about Postgres-XC version and the PostgreSQL version it is based on.

• Refactoring of pg_regress.c due to bugs with O2 compilation (Michael Paquier)

  When code was being compiled with O2, initialization and start-up of GTM was done twice, creating errors in log files. In consequence, functions related to start and initialization of PGXC nodes and GTM are unified to clarify code and additional checks are added to verify that nodes have been correctly started.

• Fix compilation flag O2 not set in CFLAGS at configure step by default (Michael Paquier)

  The flag -DPGXC is mandatory for correct compilation and is set once all the compilation flags have been determined.

• Block EXECUTE DIRECT when used from a Datanode (Michael Paquier)

  This caused the Datanode to crash, and it is however a functionality only usable by Coordinators.

• Fix snapshot management on slave nodes and on Datanodes accessed directly by applications (Michael Paquier)

  A slave node session acquires its snapshot directly from WAL like PostgreSQL does. Also, a Datanode is now able to get a global snapshot directly from GTM, avoiding WARNING messages that a snapshot is not available.

  Queries interacting with remote nodes running on slave nodes are blocked. Such queries need a global snapshot and a new transaction ID, but sessions on slave nodes cannot acquire that.

• Correct query generation of COPY with quoted table names (Nikhil Sontakke)

• Fix race condition in GXID generation for autovacuum wraparound (Michael Paquier)

  When an autovacuum wraparound begins, a global snapshot with a global GXID is obtained from GTM. After getting this snapshot, the code checks for wraparound conditions. Before this check, the lock on GXID generation is released and then taken back. Just after taking back the lock the GXID is updated to the latest value in cache. However, during this time it is possible that ShmemVariableCache->nextXid is updated by another local session that obtained a transaction ID from GTM. GXID generation concurrency is managed globally at GTM, so we cannot rely here on the local ShmemVariableCache->nextXid to update the GXID of autovacuum. Updating it to a wrong value makes is async with GTM information and such autovacuum is not able to gather a global snapshot correctly, making the autovacuum session inconsistent with the other global sessions. This could fatally crash a node under certain conditions like GXID being updated concurrently by multiple sessions at the same time during the wraparound condition checks.

• Fix for DDL ALTER ROLE ... REPLICATION not recognized (Nikhil Sontakke)

  PostgreSQL 9.1 code uses the word REPLICATION in ALTER TABLE as a portion of IDENT. However it is a reserved keyword in Postgres-XC due to its use in clause DISTRIBUTE BY, extension of CREATE TABLE and ALTER TABLE. So here what is done is completing deparsing of ALTER ROLE to use REPLICATION as a normal reserved keyword.

• Fix GTM startup with "*" set as listen_addresses on OSX (Michael Paquier)

  On OSX, GTM was not able to start with default parameter values. This had side effects on regressions with make check, making impossible to run a simple default regression test.