Informix - Problem description
Problem IT31417 | Status: Closed |
HASH JOINS LEADING TO VARIOUS FAILURES | |
product: | |
INFORMIX SERVER / 5725A3900 / C10 - IDS 12.10 | |
Problem description: | |
The following failures were reported against a special build of IDS 12.10.FC7W1 running on linux x86_64. The problem appears to be platform independent. There have been many variations of SQL that led to the failure, some complex, and some more simple queries against sysmaster. In one case the trigger seemed to be running low on temp space, but that was not always the case for reported assertions. A couple SQL examples follow that were against sysmaster database, but this database is not necessary, just the hash join operation. select count(*) from sysmaster:syssesprof where longtxs = 1 select dbsname, case when tab_name is NULL then trim(tabname) else trim(tab_name)||':'||trim(tabname) end tabname , bufreads, pagreads, bufwrites, lockreqs, seqscans, nrows from sysptprof p, sysptnhdr h, outer tabprof_tmp a where dbsname != 'sysmaster' and dbsname != 'sysutils' and tabname not in ('sysusers','sysindexes','syscolumns','TBLSpace') and p.partnum = h.partnum and trim(a.db_name) =trim(dbsname) and trim(a.idx_name)=trim(tabname) order by bufreads desc; There have been a variety of similar stacks, some being like: afstack mt_ex_throw_sig afsig_handler next_row hash_process_all_groups unordered_group_by_next filltemp scan_open materialize_viewtmp materialize_viewtmp prepselect open_cursor sql_open sq_open sqmain th_init_initgls startup afstack mt_ex_throw_sig afsig_handler search_join_hash_table build_hash_table setup_new_environment hjoin_next sort_open filltemp scan_open materialize_viewtmp materialize_viewtmp materialize_viewtmp prepselect open_cursor sql_open sq_open sqmain spawn_thread th_init_initgls startup afstack afhandler affail_interface ptmapx hash_read_buffer hash_read_tuple hash_read_tuple_from_temp next_row hash_process_all_groups unordered_group_by_next filltemp scan_open materialize_viewtmp materialize_viewtmp prepselect open_cursor sql_open sq_open sqmain spawn_thread th_init_initgls startup afstack mt_ex_throw_sig afsig_handler search_join_hash_table setup_new_environment hjoin_next join_next join_next sort_open prepselect open_cursor sql_open sq_open sqmain spawn_thread th_init_initgls startup | |
Problem Summary: | |
**************************************************************** * USERS AFFECTED: * * Users of Informix Server prior to 12.10.xC14 and 14.10.xC4. * **************************************************************** * PROBLEM DESCRIPTION: * * See Error Description * **************************************************************** * RECOMMENDATION: * * Update to Informix Server 12.10.xC14 or 14.10.xC4 (when * * available). * **************************************************************** | |
Local Fix: | |
Solution | |
Workaround | |
**************************************************************** * USERS AFFECTED: * * Users of Informix Server prior to 12.10.xC14 and 14.10.xC4. * **************************************************************** * PROBLEM DESCRIPTION: * * See Error Description * **************************************************************** * RECOMMENDATION: * * Update to Informix Server 12.10.xC14 or 14.10.xC4 (when * * available). * **************************************************************** | |
Comment | |
Fixed in Informix Server 12.10.xC14 and 14.10.xC4. The fix is disabled by default. To enable the fix set 0x10000000 in SQL_DEF_CTRL. 1) query current server SQL_DEF_CTRL value > select * from sysmaster:sysconfig where cf_name='SQL_DEF_CTRL'; cf_id 262 cf_name SQL_DEF_CTRL cf_flags 50368576 cf_original cf_effective 0x44004a0 cf_default 0x44004a0 2) turn on bit 0x10000000; in this case that would be 0x144004a0, then set it with onmode: onmode -wm SQL_DEF_CTRL=0x144004a0 | |
Timestamps | |
Date - problem reported : Date - problem closed : Date - last modified : | 06.01.2020 03.03.2020 29.05.2020 |
Problem solved at the following versions (IBM BugInfos) | |
12.10.xC14, 14.10.xC4 | |
Problem solved according to the fixlist(s) of the following version(s) |