1. 26 Jun, 2007 1 commit
    • unknown's avatar
      ndb - bug#29331 (51) · 3a4078c6
      unknown authored
          Add better handling of GCP Stop
          Only kill "offending" node
      
      
      storage/ndb/src/kernel/blocks/ERROR_codes.txt:
        add new error codes
      storage/ndb/src/kernel/blocks/dbdih/Dbdih.hpp:
        add better GCP stop handling
      storage/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp:
        add better GCP stop handling
      3a4078c6
  2. 25 Jun, 2007 4 commits
  3. 20 Jun, 2007 1 commit
  4. 18 Jun, 2007 1 commit
  5. 14 Jun, 2007 1 commit
    • unknown's avatar
      bug#29099 - slow backup for disk data · acd10b9f
      unknown authored
      - implement read ahead during disk data scan
      
      
      storage/ndb/src/kernel/blocks/dbtup/Dbtup.hpp:
        maximum read ahead
      storage/ndb/src/kernel/blocks/dbtup/DbtupGen.cpp:
        maximum read ahead should not be larger than page buffer size
        maximum aslo set to 32 pages
      storage/ndb/src/kernel/blocks/dbtup/DbtupScan.cpp:
        read ahead for disk data scan
      storage/ndb/src/kernel/blocks/pgman.cpp:
        set higher max_io_waits for read ahead
        round off upwards in page_buffer size
      storage/ndb/src/kernel/vm/SimulatedBlock.cpp:
        add null callback to be used for read ahead (no callback needed)
      storage/ndb/src/kernel/vm/SimulatedBlock.hpp:
        add null callback to be used for read ahead (no callback needed)
      storage/ndb/src/ndbapi/TransporterFacade.cpp:
        remove the debug trace, not needed any more
      storage/ndb/test/tools/hugoFill.cpp:
        allow setting database in hugoFill
      acd10b9f
  6. 12 Jun, 2007 3 commits
    • unknown's avatar
      Bug#29044 - memory buddy allocator "unoptimal" memory handling · a73a7b6c
      unknown authored
      - add config param to have better behavior with large tables
      
      a73a7b6c
    • unknown's avatar
      extend backup dump to give more info · 83e066b1
      unknown authored
      83e066b1
    • unknown's avatar
      ndb - bug#29044 · de7b2e9a
      unknown authored
        Improve buddy high order allocation
        Make removeCommonArea O(1) instead of O(N)
        Add limit to left/right search
      
      
      storage/ndb/src/kernel/blocks/dbtup/DbtupDebug.cpp:
        Add info to buddy module test about 
        1) loops being made in buddy
        2) how much was allocated
      storage/ndb/src/kernel/blocks/dbtup/DbtupPagMan.cpp:
        1) make removeCommonArea o(1) - as list is (after fix) double linked anyway
        2) set page_state = ZFREE_COMMON insertCommonArea and ~ZFREE_COMMON in removeCommonArea
        3) add max loops in search left/right
        4) add more debug info
      de7b2e9a
  7. 05 Jun, 2007 3 commits
    • unknown's avatar
      Merge tulin@bk-internal.mysql.com:/home/bk/mysql-5.1-telco-gca · 05ff34a2
      unknown authored
      into  poseidon.mysql.com:/home/tomas/mysql-5.1-telco-gca
      
      
      mysql-test/ndb/ndb_config_2_node.ini:
        Auto merged
      storage/ndb/include/mgmapi/mgmapi_config_parameters.h:
        Auto merged
      storage/ndb/src/kernel/blocks/dblqh/Dblqh.hpp:
        Auto merged
      storage/ndb/src/kernel/blocks/dblqh/DblqhInit.cpp:
        Auto merged
      storage/ndb/src/kernel/blocks/ndbcntr/NdbcntrMain.cpp:
        Auto merged
      storage/ndb/src/mgmsrv/ConfigInfo.cpp:
        Auto merged
      storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp:
        manual merge
      05ff34a2
    • unknown's avatar
      Bug #28751 Lots of memory locked in memory causes high kswapd · 96e8af01
      unknown authored
      - add odirect option for lcp+backup+redo log to lower CPU/kswapd usage
      - writing odirect removes need for kernel write buffers avoiding kswapd to kick in
      
      
      mysql-test/ndb/ndb_config_2_node.ini:
        run mysql-test-run using ODirect
      storage/ndb/include/mgmapi/mgmapi_config_parameters.h:
        add new config parameter to choose ODirect
      storage/ndb/include/ndb_global.h.in:
        specify alignment needed for odirect
      storage/ndb/src/kernel/blocks/backup/Backup.cpp:
        read odirect config param
        open LCP and Backup datafiles with odirect if specified
        insert empty padding record if odirect is used
        allocate buffers aligned to be able to use odirect
      storage/ndb/src/kernel/blocks/backup/Backup.hpp:
        odirect and padding options
      storage/ndb/src/kernel/blocks/backup/BackupFormat.hpp:
        add empty_record in file format
      storage/ndb/src/kernel/blocks/backup/BackupInit.cpp:
        read odirect config and allocate aligned
      storage/ndb/src/kernel/blocks/backup/FsBuffer.hpp:
        correct debug printouts
      storage/ndb/src/kernel/blocks/dblqh/Dblqh.hpp:
        read odirect config param and align buffers
      storage/ndb/src/kernel/blocks/dblqh/DblqhInit.cpp:
        read odirect config param and align buffers
      storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp:
        read config params and open redo log files with odirect if set
      storage/ndb/src/kernel/blocks/ndbfs/AsyncFile.cpp:
        aligned writing for odirect
        correct odirect open options with test+fallback if odirect fails
      storage/ndb/src/kernel/blocks/ndbfs/AsyncFile.hpp:
        align + odirect check
      storage/ndb/src/kernel/blocks/restore.cpp:
        restor block to ignore new lcp padding empty_record
      storage/ndb/src/kernel/vm/SimulatedBlock.cpp:
        alligend log buffer allocation for odirect
      storage/ndb/src/kernel/vm/SimulatedBlock.hpp:
        alligend log buffer allocation for odirect
      storage/ndb/src/mgmsrv/ConfigInfo.cpp:
        new config param for odirect, default false
      storage/ndb/tools/restore/Restore.cpp:
        ndb_restore to skip empty_record alignment padding in backup file
      96e8af01
    • unknown's avatar
      Bug #28899 not possible to set separate watchdog timeout at startup · 13f4ac11
      unknown authored
      
      storage/ndb/include/mgmapi/mgmapi_config_parameters.h:
        add new configuration parameter TimeBetweenWatchDogCheckInitial
      storage/ndb/include/portlib/NdbTick.h:
        enable timing code
      storage/ndb/src/common/portlib/NdbTick.c:
        enable timing code
      storage/ndb/src/kernel/blocks/ndbcntr/NdbcntrMain.cpp:
        read watchdog timeout to set it after malloc
      storage/ndb/src/kernel/vm/Configuration.cpp:
        read initial watchdog timeout and set it in the beginning
      storage/ndb/src/kernel/vm/Configuration.hpp:
        read initial watchdog timeout and set it in the beginning
      storage/ndb/src/kernel/vm/SimulatedBlock.cpp:
        introduce new state for "action" malloc of memory
      storage/ndb/src/kernel/vm/SimulatedBlock.hpp:
        introduce new state for "action" malloc of memory
      storage/ndb/src/kernel/vm/WatchDog.cpp:
        rewrite watchdog to check every 100ms for being stuch, but keep shutdown after 3 * interval
        for "action" == 9 (malloc)  keep old behavior and only output every interval
      storage/ndb/src/mgmsrv/ConfigInfo.cpp:
        add new configuration parameter TimeBetweenWatchDogCheckInitial
      13f4ac11
  8. 04 Jun, 2007 2 commits
    • unknown's avatar
      ndb - bug#28726 · c95fb65e
      unknown authored
        make sure to remove LCP files aswell if specifying --initial
      
      
      storage/ndb/src/kernel/blocks/ndbcntr/NdbcntrMain.cpp:
        Add removal of LCP/X directories
      c95fb65e
    • unknown's avatar
      ndb - make size of redo log files (fragment log files) configurable · 4979bfc9
      unknown authored
        using new config variable FragmentLogFileSize (4M - 1G)
      
      
      mysql-test/ndb/ndb_config_1_node.ini:
        change log file size (test)
      mysql-test/ndb/ndb_config_2_node.ini:
        change log file size (test)
      storage/ndb/include/mgmapi/mgmapi_config_parameters.h:
        add new confif parameter
      storage/ndb/src/kernel/blocks/dblqh/Dblqh.hpp:
        make logfile size configurable
      storage/ndb/src/kernel/blocks/dblqh/DblqhInit.cpp:
        make logfile size configurable
      storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp:
        make logfile size configurable
      storage/ndb/src/mgmsrv/ConfigInfo.cpp:
        add new config variable
      storage/ndb/src/ndbapi/ndberror.c:
        update error message
      4979bfc9
  9. 03 Jun, 2007 1 commit
    • unknown's avatar
      Bug#20612. · 1b5c8df7
      unknown authored
      
      storage/ndb/src/kernel/blocks/pgman.cpp:
        a) in one case pl_queue.remove(ptr) was not followed
        by state &= ~ Page_entry::ONQUEUE.
        
        b) when collecting initial hot entries have to remove
        from queue if somehow got there.
        
        b) is easy to get with large buffer cache (256M).
        a) or b) is probably cause of bug#20612.
      1b5c8df7
  10. 30 May, 2007 4 commits
  11. 29 May, 2007 6 commits
    • unknown's avatar
      shorten some files for tar to work · bb82fa3b
      unknown authored
      
      storage/ndb/ndbapi-examples/ndbapi_simple_index/main.cpp:
        Rename: storage/ndb/ndbapi-examples/ndbapi_simple_index/ndbapi_simple_index.cpp -> storage/ndb/ndbapi-examples/ndbapi_simple_index/main.cpp
      storage/ndb/ndbapi-examples/mgmapi_logevent/main.cpp:
        Rename: storage/ndb/ndbapi-examples/mgmapi_logevent/mgmapi_logevent.cpp -> storage/ndb/ndbapi-examples/mgmapi_logevent/main.cpp
      storage/ndb/ndbapi-examples/mgmapi_logevent2/main.cpp:
        Rename: storage/ndb/ndbapi-examples/mgmapi_logevent2/mgmapi_logevent2.cpp -> storage/ndb/ndbapi-examples/mgmapi_logevent2/main.cpp
      storage/ndb/src/kernel/blocks/dblqh/redoLogReader/reader.cpp:
        Rename: storage/ndb/src/kernel/blocks/dblqh/redoLogReader/redoLogFileReader.cpp -> storage/ndb/src/kernel/blocks/dblqh/redoLogReader/reader.cpp
      storage/ndb/ndbapi-examples/ndbapi_simple_dual/main.cpp:
        Rename: storage/ndb/ndbapi-examples/ndbapi_simple_dual/ndbapi_simple_dual.cpp -> storage/ndb/ndbapi-examples/ndbapi_simple_dual/main.cpp
      bb82fa3b
    • unknown's avatar
      Merge tulin@bk-internal.mysql.com:/home/bk/mysql-5.1-telco-gca · 6a8a7f7d
      unknown authored
      into  whalegate.ndb.mysql.com:/home/tomas/mysql-5.1-telco-gca
      
      6a8a7f7d
    • unknown's avatar
      Merge joreland@bk-internal.mysql.com:/home/bk/mysql-5.1-telco-gca · 7eb2b153
      unknown authored
      into  perch.ndb.mysql.com:/home/jonas/src/51-telco-gca
      
      
      storage/ndb/src/kernel/blocks/dblqh/Dblqh.hpp:
        Auto merged
      storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp:
        Auto merged
      storage/ndb/src/kernel/blocks/dbtup/DbtupScan.cpp:
        Auto merged
      storage/ndb/test/run-test/daily-basic-tests.txt:
        Auto merged
      7eb2b153
    • unknown's avatar
      Merge perch.ndb.mysql.com:/home/jonas/src/50-work · 71048258
      unknown authored
      into  perch.ndb.mysql.com:/home/jonas/src/51-telco-gca
      
      
      storage/ndb/test/run-test/daily-basic-tests.txt:
        merge
      71048258
    • unknown's avatar
      Merge perch.ndb.mysql.com:/home/jonas/src/41-work · f863fc2a
      unknown authored
      into  perch.ndb.mysql.com:/home/jonas/src/50-work
      
      
      ndb/test/run-test/daily-basic-tests.txt:
        merge
      f863fc2a
    • unknown's avatar
      ndb - fix typo · 497d0e58
      unknown authored
      497d0e58
  12. 28 May, 2007 6 commits
  13. 25 May, 2007 1 commit
    • unknown's avatar
      Bug #28525 Node failures in PGMAN at ndbrequire (line 430) · 3608d311
      unknown authored
      
      storage/ndb/src/kernel/blocks/pgman.cpp:
        Under heavy insert PGMAN can run out of page entries
        even when set to 100 times page cache entries.
        
        In this use pattern the extra entries remain idle
        on LIRS stack.  Only ONSTACK is set.  There is not
        enough activity to free them the normal way.
        
        A study of PGMAN / DBTUP behaviour is needed.
        
        This patch adds new sublist SL_IDLE.  When page entry
        pool is empty, an idle entry is released from SL_IDLE
        front if there is any.  Otherwise, we still crash.
        
        The factor above is set from 100 to 10 (still high).
      storage/ndb/src/kernel/blocks/pgman.hpp:
        Under heavy insert PGMAN can run out of page entries
        even when set to 100 times page cache entries.
        
        In this use pattern the extra entries remain idle
        on LIRS stack.  Only ONSTACK is set.  There is not
        enough activity to free them the normal way.
        
        A study of PGMAN / DBTUP behaviour is needed.
        
        This patch adds new sublist SL_IDLE.  When page entry
        pool is empty, an idle entry is released from SL_IDLE
        front if there is any.  Otherwise, we still crash.
        
        The factor above is set from 100 to 10 (still high).
      3608d311
  14. 24 May, 2007 2 commits
  15. 22 May, 2007 1 commit
    • unknown's avatar
      Bug #28593 cluster backup scans in acc index order, bad for disk data · 45cf4639
      unknown authored
      - change to scan in tup and disk order (if applicable)
      
      
      storage/ndb/src/kernel/blocks/backup/Backup.cpp:
        tup scan for backup also, not only lcp
      storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp:
        choose disk order scan for disktables during backup
        + move bug workaround for 27776 from tup to lqh
      storage/ndb/src/kernel/blocks/dbtup/DbtupDiskAlloc.cpp:
        add missing jamEntry
      storage/ndb/src/kernel/blocks/dbtup/DbtupScan.cpp:
        move bug workaround for 27776 from tup to lqh
      storage/ndb/src/kernel/blocks/pgman.cpp:
        missing jamEntry
      45cf4639
  16. 18 May, 2007 3 commits