Join Firebird!

Join Firebird Foundation to support Firebird SQL development and receive multiple bonuses

Follow Us

Select your media preference

Newsletter

Subscribe to Firebird’s Newsletter to receive the latest news

Developer's Report: QA Stress Testing & Tools Development
November 2014 - January 2015

November 2014
  • Improvements in OLTP-EMUL batch files (parsing of oltp_config.NN files)
  • Implement scripts for running OLTP-EMUL under Linux
  • Run tests for gathering info about FB performance degradation under load (20-200 connections, 360 minutes, FW=ON, v3 SS)
  • Researched concurrent performance with reverted IDs as PK
  • Attempted to to catch error 335545060 ("Missing security context") using special build prepared by Alex
  • Run heavy OLTP-EMUL test: compare performance for FW=ON vs OFF (v3 SS)
  • Investigated performance when database is stored on RAM drive (/dev/shm)
  • Found very slow connection establishing while bulk inserting into table with many 'heavy' indices (depth = 10-12, long varchar fields, page size = 16KB)
  • Investigated CORE-4416 (table with too wide index keys)
 
December 2014
  • Unable to connect to database when DML with multiple delete+insert statements is running (CORE-4632)
  • Tested fix for CORE-4571 (ICU restore after copying database with UTF8 data from windows to linux host and vice versa)
  • Found firebird.log being polluted with "setsockopt: error setting IPV6_V6ONLY to 0" on Windows XP
  • Attempted to compare U+0418 and U+0419 (cyrillic `I` and `Short I`) using accent insensitive collation: these letters are considered equal but in fact they differ
  • Validated that no bitmap is created for queries with "where exists (select * from T where ... ORDER BY )" when = PK / UK
  • OLTP-EMUL: remove PK constraints from tables which actually do not need them at all
  • Found memory leak in v3 under brute-force attack (CORE-4643)
  • Catched errors "335545060 : Missing security context " and "335544344 : I/O error during "open" operation" for existing database under heavy concurrent DML (CORE-4644)
  • Crash on infinite mutual SP calls instead of error "Too many concurrent executions of the same request" (CORE-4653)
  • Started to learn the FBTEST framework

January 2015
  • OLTP-EMUL: Server hangs when .fdb *and* FIREBIRD_LOCK are on /dev/shm
  • Found poor performance of FULL JOIN of table itself (4 times, on PK field; 100 rows in table) in comparison with MS SQL 2005 XE
  • Found new bug in MERGE statement when INDEX present
  • Misc OLTP-EMUL improvements
  • OLTP-EMUL, 100 connections: several strange messages in firebird.log (some phrases from PSQL call stack, alerts about unable to read from GTT file etc, "page type 4 lock conversion denied"; /O error during "read" operation for file "/var/db/fb30/oltp30.fdb")
  • Wrong result when use "where STARTING WITH <:value> ORDER BY ' and field_C is leading part of compound index key: { field_C, field_N } (CORE-4665)
  • Run OLTP-EMUL on special build provided by Vlad in order to find conditions when 'page N type M lock denied'
  • Select from mon$table_stats doesn`t work on SC and CS (CORE-4668)
  • Validated that CORE-4384 is really fixed (QA test is impossible due to huge database size)
  • OLTP-EMUL: separate branch for investigating how 'pseudo-partitioning' affects the performance
  • Implemented QA regression tests for many old bugs
 
Pavel Zotov
Moscow, Russian Federation
November 2014 - January 2015

November 2014
  • Improvements in OLTP-EMUL batch files (parsing of oltp_config.NN files)
  • Implement scripts for running OLTP-EMUL under Linux
  • Run tests for gathering info about FB performance degradation under load (20-200 connections, 360 minutes, FW=ON, v3 SS)
  • Researched concurrent performance with reverted IDs as PK
  • Attempted to to catch error 335545060 ("Missing security context") using special build prepared by Alex
  • Run heavy OLTP-EMUL test: compare performance for FW=ON vs OFF (v3 SS)
  • Investigated performance when database is stored on RAM drive (/dev/shm)
  • Found very slow connection establishing while bulk inserting into table with many 'heavy' indices (depth = 10-12, long varchar fields, page size = 16KB)
  • Investigated CORE-4416 (table with too wide index keys)
 
December 2014
  • Unable to connect to database when DML with multiple delete+insert statements is running (CORE-4632)
  • Tested fix for CORE-4571 (ICU restore after copying database with UTF8 data from windows to linux host and vice versa)
  • Found firebird.log being polluted with "setsockopt: error setting IPV6_V6ONLY to 0" on Windows XP
  • Attempted to compare U+0418 and U+0419 (cyrillic `I` and `Short I`) using accent insensitive collation: these letters are considered equal but in fact they differ
  • Validated that no bitmap is created for queries with "where exists (select * from T where ... ORDER BY )" when = PK / UK
  • OLTP-EMUL: remove PK constraints from tables which actually do not need them at all
  • Found memory leak in v3 under brute-force attack (CORE-4643)
  • Catched errors "335545060 : Missing security context " and "335544344 : I/O error during "open" operation" for existing database under heavy concurrent DML (CORE-4644)
  • Crash on infinite mutual SP calls instead of error "Too many concurrent executions of the same request" (CORE-4653)
  • Started to learn the FBTEST framework

January 2015
  • OLTP-EMUL: Server hangs when .fdb *and* FIREBIRD_LOCK are on /dev/shm
  • Found poor performance of FULL JOIN of table itself (4 times, on PK field; 100 rows in table) in comparison with MS SQL 2005 XE
  • Found new bug in MERGE statement when INDEX present
  • Misc OLTP-EMUL improvements
  • OLTP-EMUL, 100 connections: several strange messages in firebird.log (some phrases from PSQL call stack, alerts about unable to read from GTT file etc, "page type 4 lock conversion denied"; /O error during "read" operation for file "/var/db/fb30/oltp30.fdb")
  • Wrong result when use "where STARTING WITH <:value> ORDER BY ' and field_C is leading part of compound index key: { field_C, field_N } (CORE-4665)
  • Run OLTP-EMUL on special build provided by Vlad in order to find conditions when 'page N type M lock denied'
  • Select from mon$table_stats doesn`t work on SC and CS (CORE-4668)
  • Validated that CORE-4384 is really fixed (QA test is impossible due to huge database size)
  • OLTP-EMUL: separate branch for investigating how 'pseudo-partitioning' affects the performance
  • Implemented QA regression tests for many old bugs
 
Pavel Zotov
Moscow, Russian Federation