Someone have a technique to alter tables with FK ?

Hi, 11.50 FC9X6 , AIX 6.1 We suffer with one situation at our production where just not found a acceptable way to work yet... I already open a PMR (for me this is a defect) but for IBM support is just "work as design" .... so, no solution... Just a correction / better documentation was made. THE PROBLEM - When I alter one table (any changes cited bellow), no matter if I get the exclusive lock for this table and all others what exists any FK from/to it... new users sessions, when try update/insert/delete any for this references tables get the error -710 Table <table-name> has been dropped, altered, or renamed. - create/drop index online/offline OR add/drop trigger OR add/modify/drop column OR alter any constraint... TODAY, THE UNIQUE SOLUTION WHEN THIS OCCUR: - Stop all our production, kill all users (put the instance in quiescent) and back online immediately . HOW I PROCEED WHEN NEED TO CHANGE SOME TABLE: - set IFX_DIRTY_WAIT - open dbaccess , start new transaction (begin work) - lock table in exclusive mode - grant a dummy select over the table (lock sys*) - At other session, kill all users what is accessing this table (identifying with onstat -g opn + onmode -z) - execute the alter over the table - commit If the table have any FK from/to it , I lock and kill users for this tables too. But, at 90% of the situations , after all executed successfully , the system stuck with -710 error. Do not matter if the user close all application and open a new session, the error -710 persist. THE ENVIRONMENT - IFX 11.50 FC9X6 , AIX 6.1 , 4GL 7.32 - Is a centralized 24x7 system what control all plants of the company . - Avg of 1800 sessions concurrent at high peak time and 400 sessions at lower peak time. - User sessions: 70% incoming from 4GL system (most important and where suffer with the problem) - User sessions: 25% incoming from Web/Java system - User sessions: 5% others.... - Database have AUTO_REPREPARE and STMT_CACHE active - The isolation used is 99% 'dirty read' . - The 4GL system work mostly with prepared SQLs where they prepare at beginner of 4GL connection then keep this prepare active during they execution. CONSIDERATIONS - If check with onstat -g dic , always have a refcnt and dirty over the tables involved. This before , during and after the alter . Even during the alter, consider the table in exclusive lock + lock over sys* (grant select). - After the alter, appear new references on onstat -g dic. (with dirty and without dirty) - After finish all alters, if the user got erro -710, close all system , open it again e try again, they stuck at -710 again and again and again and again.... - I try simulate the problem, but with few users, this behave do not occur. - Bellow is a stack trace (trapped with onmode -I 710) when the user tryed insert a register over a table where is relationed with the table altered (in this case, added a new index online over columns what not involve the FK) 0x00000001000b76a8 (oninit)afstack 0x00000001000b97bc (oninit)afhandler 0x0000000100130b90 (oninit)check_traperror 0x00000001001faa94 (oninit)sqerr 0x00000001001fa550 (oninit)sqerr1 0x00000001001f9e54 (oninit)sqnameerr 0x000000010066e4dc (oninit)chkmajvers 0x000000010066eb24 (oninit)openrel 0x00000001006523b4 (oninit)chkparent 0x000000010065b2d0 (oninit)chkrowcons 0x0000000100782790 (oninit)addone 0x0000000100784e10 (oninit)insone_next 0x0000000100769994 (oninit)doinsert 0x000000010041cff8 (oninit)aud_doinsert 0x00000001004259f8 (oninit)excommand 0x000000010044b3bc (oninit)sq_execute 0x000000010022463c (oninit)sqmain 0x000000010037c160 (oninit)listen_verify 0x000000010037a608 (oninit)spawn_thread 0x0000000100dd528c (oninit)startup At my point of view, Informix isn't able to work 100% online if you need to change your scheme (add a simple index using online mode) when you use referential integrity. I believed this is because our method of work (prepare at start of application, and keep this prepared active to reuse it), anyway, should we suffer with this consequences!??? Is there some technique to alter a table what I miss here ? Cesar

Long post... Error 710 should only happen on PREPARED statements... And the applications should know how to handle the -710 error (re-prepare the statement) If you're hitting -710 without a prepared statement then this needs further investigation. If -710 comes only from PREPARED statements, then AUTO_REPREPARE should help, but it's not guaranteed 100% of time. But you say that in a new session will also get the -710. This doesn't look normal... Have you tried to flush the statement cache? Can you give the PMR number? Do you have a test case? Have you tried it on a different version? At first glance it doesn't look good, but I'd need to look into it very carefully before issuing a definitive answer... specially if you already checked with support... Regards. On Wed, May 2, 2012 at 2:38 PM, Cesar Inacio Martins < cesar_inacio_martins@yahoo.com.br> wrote: > Hi, > > 11.50 FC9X6 , AIX 6.1 > > We suffer with one situation at our production where just not found a > acceptable way to work yet... > I already open a PMR (for me this is a defect) but for IBM support is > just "work as design" .... so, no solution... > Just a correction / better documentation was made. > > THE PROBLEM > - When I alter one table (any changes cited bellow), no matter if I get > the exclusive lock for this table and all others what exists any FK > from/to it... new users sessions, when try update/insert/delete any for > this references tables get the error -710 Table <table-name> has been > dropped, altered, or renamed. > - create/drop index online/offline OR add/drop trigger OR > add/modify/drop column OR alter any constraint... > > TODAY, THE UNIQUE SOLUTION WHEN THIS OCCUR: > - Stop all our production, kill all users (put the instance in > quiescent) and back online immediately . > > HOW I PROCEED WHEN NEED TO CHANGE SOME TABLE: > - set IFX_DIRTY_WAIT > - open dbaccess , start new transaction (begin work) > - lock table in exclusive mode > - grant a dummy select over the table (lock sys*) > - At other session, kill all users what is accessing this table > (identifying with onstat -g opn + onmode -z) > - execute the alter over the table > - commit > If the table have any FK from/to it , I lock and kill users for this > tables too. > But, at 90% of the situations , after all executed successfully , the > system stuck with -710 error. > Do not matter if the user close all application and open a new session, > the error -710 persist. > > THE ENVIRONMENT > - IFX 11.50 FC9X6 , AIX 6.1 , 4GL 7.32 > - Is a centralized 24x7 system what control all plants of the company . > - Avg of 1800 sessions concurrent at high peak time and 400 sessions at > lower peak time. > - User sessions: 70% incoming from 4GL system (most important and where > suffer with the problem) > - User sessions: 25% incoming from Web/Java system > - User sessions: 5% others.... > - Database have AUTO_REPREPARE and STMT_CACHE active > - The isolation used is 99% 'dirty read' . > - The 4GL system work mostly with prepared SQLs where they prepare at > beginner of 4GL connection then keep this prepare active during they > execution. > > CONSIDERATIONS > - If check with onstat -g dic , always have a refcnt and dirty over the > tables involved. This before , during and after the alter . Even during > the alter, consider the table in exclusive lock + lock over sys* (grant > select). > - After the alter, appear new references on onstat -g dic. (with dirty > and without dirty) > - After finish all alters, if the user got erro -710, close all system , > open it again e try again, they stuck at -710 again and again and again > and again.... > - I try simulate the problem, but with few users, this behave do not occur. > - Bellow is a stack trace (trapped with onmode -I 710) when the user > tryed insert a register over a table where is relationed with the table > altered (in this case, added a new index online over columns what not > involve the FK) > > 0x00000001000b76a8 (oninit)afstack > > 0x00000001000b97bc (oninit)afhandler > > 0x0000000100130b90 (oninit)check_traperror > > 0x00000001001faa94 (oninit)sqerr > > 0x00000001001fa550 (oninit)sqerr1 > > 0x00000001001f9e54 (oninit)sqnameerr > > 0x000000010066e4dc (oninit)chkmajvers > > 0x000000010066eb24 (oninit)openrel > > 0x00000001006523b4 (oninit)chkparent > > 0x000000010065b2d0 (oninit)chkrowcons > > 0x0000000100782790 (oninit)addone > > 0x0000000100784e10 (oninit)insone_next > > 0x0000000100769994 (oninit)doinsert > > 0x000000010041cff8 (oninit)aud_doinsert > > 0x00000001004259f8 (oninit)excommand > > 0x000000010044b3bc (oninit)sq_execute > > 0x000000010022463c (oninit)sqmain > > 0x000000010037c160 (oninit)listen_verify > > 0x000000010037a608 (oninit)spawn_thread > > 0x0000000100dd528c (oninit)startup > > At my point of view, Informix isn't able to work 100% online if you need > to change your scheme (add a simple index using online mode) when you > use referential integrity. > I believed this is because our method of work (prepare at start of > application, and keep this prepared active to reuse it), anyway, should > we suffer with this consequences!??? > > Is there some technique to alter a table what I miss here ? > > Cesar > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > -- Fernando Nunes Portugal http://informix-technology.blogspot.com My email works... but I don't check it frequently... --00235452f64c14f90d04bf0f4faf

Hmmmm.... was the dictionary cache mentioned as a possible cause for this? I'd like to have the ability to flush all/partial entries in any of the engine caches... I've faced a couple of situations where this would help (bugs found and fixed, but meanwhile it would have been nice). AFAIK there is currently no feature request for this. Regards. On Wed, May 2, 2012 at 4:26 PM, Fernando Nunes <domusonline@gmail.com>wrote: > Long post... > Error 710 should only happen on PREPARED statements... And the applications > should know how to handle the -710 error (re-prepare the statement) > If you're hitting -710 without a prepared statement then this needs further > investigation. > If -710 comes only from PREPARED statements, then AUTO_REPREPARE should > help, but it's not guaranteed 100% of time. > But you say that in a new session will also get the -710. This doesn't look > normal... Have you tried to flush the statement cache? > > Can you give the PMR number? Do you have a test case? Have you tried it on > a different version? > > At first glance it doesn't look good, but I'd need to look into it very > carefully before issuing a definitive answer... specially if you already > checked with support... > > Regards. > > On Wed, May 2, 2012 at 2:38 PM, Cesar Inacio Martins < > cesar_inacio_martins@yahoo.com.br> wrote: > > > Hi, > > > > 11.50 FC9X6 , AIX 6.1 > > > > We suffer with one situation at our production where just not found a > > acceptable way to work yet... > > I already open a PMR (for me this is a defect) but for IBM support is > > just "work as design" .... so, no solution... > > Just a correction / better documentation was made. > > > > THE PROBLEM > > - When I alter one table (any changes cited bellow), no matter if I get > > the exclusive lock for this table and all others what exists any FK > > from/to it... new users sessions, when try update/insert/delete any for > > this references tables get the error -710 Table <table-name> has been > > dropped, altered, or renamed. > > - create/drop index online/offline OR add/drop trigger OR > > add/modify/drop column OR alter any constraint... > > > > TODAY, THE UNIQUE SOLUTION WHEN THIS OCCUR: > > - Stop all our production, kill all users (put the instance in > > quiescent) and back online immediately . > > > > HOW I PROCEED WHEN NEED TO CHANGE SOME TABLE: > > - set IFX_DIRTY_WAIT > > - open dbaccess , start new transaction (begin work) > > - lock table in exclusive mode > > - grant a dummy select over the table (lock sys*) > > - At other session, kill all users what is accessing this table > > (identifying with onstat -g opn + onmode -z) > > - execute the alter over the table > > - commit > > If the table have any FK from/to it , I lock and kill users for this > > tables too. > > But, at 90% of the situations , after all executed successfully , the > > system stuck with -710 error. > > Do not matter if the user close all application and open a new session, > > the error -710 persist. > > > > THE ENVIRONMENT > > - IFX 11.50 FC9X6 , AIX 6.1 , 4GL 7.32 > > - Is a centralized 24x7 system what control all plants of the company . > > - Avg of 1800 sessions concurrent at high peak time and 400 sessions at > > lower peak time. > > - User sessions: 70% incoming from 4GL system (most important and where > > suffer with the problem) > > - User sessions: 25% incoming from Web/Java system > > - User sessions: 5% others.... > > - Database have AUTO_REPREPARE and STMT_CACHE active > > - The isolation used is 99% 'dirty read' . > > - The 4GL system work mostly with prepared SQLs where they prepare at > > beginner of 4GL connection then keep this prepare active during they > > execution. > > > > CONSIDERATIONS > > - If check with onstat -g dic , always have a refcnt and dirty over the > > tables involved. This before , during and after the alter . Even during > > the alter, consider the table in exclusive lock + lock over sys* (grant > > select). > > - After the alter, appear new references on onstat -g dic. (with dirty > > and without dirty) > > - After finish all alters, if the user got erro -710, close all system , > > open it again e try again, they stuck at -710 again and again and again > > and again.... > > - I try simulate the problem, but with few users, this behave do not > occur. > > - Bellow is a stack trace (trapped with onmode -I 710) when the user > > tryed insert a register over a table where is relationed with the table > > altered (in this case, added a new index online over columns what not > > involve the FK) > > > > 0x00000001000b76a8 (oninit)afstack > > > > 0x00000001000b97bc (oninit)afhandler > > > > 0x0000000100130b90 (oninit)check_traperror > > > > 0x00000001001faa94 (oninit)sqerr > > > > 0x00000001001fa550 (oninit)sqerr1 > > > > 0x00000001001f9e54 (oninit)sqnameerr > > > > 0x000000010066e4dc (oninit)chkmajvers > > > > 0x000000010066eb24 (oninit)openrel > > > > 0x00000001006523b4 (oninit)chkparent > > > > 0x000000010065b2d0 (oninit)chkrowcons > > > > 0x0000000100782790 (oninit)addone > > > > 0x0000000100784e10 (oninit)insone_next > > > > 0x0000000100769994 (oninit)doinsert > > > > 0x000000010041cff8 (oninit)aud_doinsert > > > > 0x00000001004259f8 (oninit)excommand > > > > 0x000000010044b3bc (oninit)sq_execute > > > > 0x000000010022463c (oninit)sqmain > > > > 0x000000010037c160 (oninit)listen_verify > > > > 0x000000010037a608 (oninit)spawn_thread > > > > 0x0000000100dd528c (oninit)startup > > > > At my point of view, Informix isn't able to work 100% online if you need > > to change your scheme (add a simple index using online mode) when you > > use referential integrity. > > I believed this is because our method of work (prepare at start of > > application, and keep this prepared active to reuse it), anyway, should > > we suffer with this consequences!??? > > > > Is there some technique to alter a table what I miss here ? > > > > Cesar > > > > > > > > > > ******************************************************************************* > > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > > > -- > Fernando Nunes > Portugal > > http://informix-technology.blogspot.com > My email works... but I don't check it frequently... > > --00235452f64c14f90d04bf0f4faf > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > -- Fernando Nunes Portugal http://informix-technology.blogspot.com My email works... but I don't check it frequently... --20cf300faeab52a6b904bf0f6149

Hi Fernando, We expected few of -710 erros because we don't handle it at our applications. When an user get this error, the application show a message and close automatically forcing the user open a new session, when the user try run the application again the problem persists... I try create a test case, this behave don't occur. I don't have how simulate 400 sessions concurrent, only fews sessions and all works fine, the error -710 don't persist. The version what we open the PMR and try figure out why this occur is 11.50 FC8 , and with FC9 and FC9X6 still occur. I don't try with other version because I can't able to simulate or create a test case, so occur only at our production. The statement cache, when I open the first PMR it was disabled because of other problem with it (STMT_CACHE = 1 ), we enable it (STMT_CACHE =2 ) when we update to FC9 or FC9X6 (I don't remember with sure what of both). Honestly I don't remember if I try flush the stat_cache. PMR 20357,228,631 PMR 23072,228,631 Fernando, if you need more information, please ask me . We need solve this behave or discovery a way to work around because affect to much the company. Thanks Cesar On 2/5/2012 12:26, Fernando Nunes wrote: > Long post... > Error 710 should only happen on PREPARED statements... And the applications > should know how to handle the -710 error (re-prepare the statement) > If you're hitting -710 without a prepared statement then this needs further > investigation. > If -710 comes only from PREPARED statements, then AUTO_REPREPARE should > help, but it's not guaranteed 100% of time. > But you say that in a new session will also get the -710. This doesn't look > normal... Have you tried to flush the statement cache? > > Can you give the PMR number? Do you have a test case? Have you tried it on > a different version? > > At first glance it doesn't look good, but I'd need to look into it very > carefully before issuing a definitive answer... specially if you already > checked with support... > > Regards. > > On Wed, May 2, 2012 at 2:38 PM, Cesar Inacio Martins< > cesar_inacio_martins@yahoo.com.br> wrote: > >> Hi, >> >> 11.50 FC9X6 , AIX 6.1 >> >> We suffer with one situation at our production where just not found a >> acceptable way to work yet... >> I already open a PMR (for me this is a defect) but for IBM support is >> just "work as design" .... so, no solution... >> Just a correction / better documentation was made. >> >> THE PROBLEM >> - When I alter one table (any changes cited bellow), no matter if I get >> the exclusive lock for this table and all others what exists any FK >> from/to it... new users sessions, when try update/insert/delete any for >> this references tables get the error -710 Table<table-name> has been >> dropped, altered, or renamed. >> - create/drop index online/offline OR add/drop trigger OR >> add/modify/drop column OR alter any constraint... >> >> TODAY, THE UNIQUE SOLUTION WHEN THIS OCCUR: >> - Stop all our production, kill all users (put the instance in >> quiescent) and back online immediately . >> >> HOW I PROCEED WHEN NEED TO CHANGE SOME TABLE: >> - set IFX_DIRTY_WAIT >> - open dbaccess , start new transaction (begin work) >> - lock table in exclusive mode >> - grant a dummy select over the table (lock sys*) >> - At other session, kill all users what is accessing this table >> (identifying with onstat -g opn + onmode -z) >> - execute the alter over the table >> - commit >> If the table have any FK from/to it , I lock and kill users for this >> tables too. >> But, at 90% of the situations , after all executed successfully , the >> system stuck with -710 error. >> Do not matter if the user close all application and open a new session, >> the error -710 persist. >> >> THE ENVIRONMENT >> - IFX 11.50 FC9X6 , AIX 6.1 , 4GL 7.32 >> - Is a centralized 24x7 system what control all plants of the company . >> - Avg of 1800 sessions concurrent at high peak time and 400 sessions at >> lower peak time. >> - User sessions: 70% incoming from 4GL system (most important and where >> suffer with the problem) >> - User sessions: 25% incoming from Web/Java system >> - User sessions: 5% others.... >> - Database have AUTO_REPREPARE and STMT_CACHE active >> - The isolation used is 99% 'dirty read' . >> - The 4GL system work mostly with prepared SQLs where they prepare at >> beginner of 4GL connection then keep this prepare active during they >> execution. >> >> CONSIDERATIONS >> - If check with onstat -g dic , always have a refcnt and dirty over the >> tables involved. This before , during and after the alter . Even during >> the alter, consider the table in exclusive lock + lock over sys* (grant >> select). >> - After the alter, appear new references on onstat -g dic. (with dirty >> and without dirty) >> - After finish all alters, if the user got erro -710, close all system , >> open it again e try again, they stuck at -710 again and again and again >> and again.... >> - I try simulate the problem, but with few users, this behave do not occur. >> - Bellow is a stack trace (trapped with onmode -I 710) when the user >> tryed insert a register over a table where is relationed with the table >> altered (in this case, added a new index online over columns what not >> involve the FK) >> >> 0x00000001000b76a8 (oninit)afstack >> >> 0x00000001000b97bc (oninit)afhandler >> >> 0x0000000100130b90 (oninit)check_traperror >> >> 0x00000001001faa94 (oninit)sqerr >> >> 0x00000001001fa550 (oninit)sqerr1 >> >> 0x00000001001f9e54 (oninit)sqnameerr >> >> 0x000000010066e4dc (oninit)chkmajvers >> >> 0x000000010066eb24 (oninit)openrel >> >> 0x00000001006523b4 (oninit)chkparent >> >> 0x000000010065b2d0 (oninit)chkrowcons >> >> 0x0000000100782790 (oninit)addone >> >> 0x0000000100784e10 (oninit)insone_next >> >> 0x0000000100769994 (oninit)doinsert >> >> 0x000000010041cff8 (oninit)aud_doinsert >> >> 0x00000001004259f8 (oninit)excommand >> >> 0x000000010044b3bc (oninit)sq_execute >> >> 0x000000010022463c (oninit)sqmain >> >> 0x000000010037c160 (oninit)listen_verify >> >> 0x000000010037a608 (oninit)spawn_thread >> >> 0x0000000100dd528c (oninit)startup >> >> At my point of view, Informix isn't able to work 100% online if you need >> to change your scheme (add a simple index using online mode) when you >> use referential integrity. >> I believed this is because our method of work (prepare at start of >> application, and keep this prepared active to reuse it), anyway, should >> we suffer with this consequences!??? >> >> Is there some technique to alter a table what I miss here ? >> >> Cesar >> >> >> >> > ******************************************************************************* >> Forum Note: Use "Reply" to post a response in the discussion forum. >> >>

If I'm not confusing anything, one of answers from support says the -710 persists because exists references to table where have dirty = yes. (onstat -g dic) or something like that. On 2/5/2012 12:31, Fernando Nunes wrote: > Hmmmm.... was the dictionary cache mentioned as a possible cause for this? > I'd like to have the ability to flush all/partial entries in any of the > engine caches... I've faced a couple of situations where this would help > (bugs found and fixed, but meanwhile it would have been nice). AFAIK there > is currently no feature request for this. > > Regards. > > On Wed, May 2, 2012 at 4:26 PM, Fernando Nunes<domusonline@gmail.com>wrote: > >> Long post... >> Error 710 should only happen on PREPARED statements... And the applications >> should know how to handle the -710 error (re-prepare the statement) >> If you're hitting -710 without a prepared statement then this needs further >> investigation. >> If -710 comes only from PREPARED statements, then AUTO_REPREPARE should >> help, but it's not guaranteed 100% of time. >> But you say that in a new session will also get the -710. This doesn't look >> normal... Have you tried to flush the statement cache? >> >> Can you give the PMR number? Do you have a test case? Have you tried it on >> a different version? >> >> At first glance it doesn't look good, but I'd need to look into it very >> carefully before issuing a definitive answer... specially if you already >> checked with support... >> >> Regards. >> >> On Wed, May 2, 2012 at 2:38 PM, Cesar Inacio Martins< >> cesar_inacio_martins@yahoo.com.br> wrote: >> >>> Hi, >>> >>> 11.50 FC9X6 , AIX 6.1 >>> >>> We suffer with one situation at our production where just not found a >>> acceptable way to work yet... >>> I already open a PMR (for me this is a defect) but for IBM support is >>> just "work as design" .... so, no solution... >>> Just a correction / better documentation was made. >>> >>> THE PROBLEM >>> - When I alter one table (any changes cited bellow), no matter if I get >>> the exclusive lock for this table and all others what exists any FK >>> from/to it... new users sessions, when try update/insert/delete any for >>> this references tables get the error -710 Table<table-name> has been >>> dropped, altered, or renamed. >>> - create/drop index online/offline OR add/drop trigger OR >>> add/modify/drop column OR alter any constraint... >>> >>> TODAY, THE UNIQUE SOLUTION WHEN THIS OCCUR: >>> - Stop all our production, kill all users (put the instance in >>> quiescent) and back online immediately . >>> >>> HOW I PROCEED WHEN NEED TO CHANGE SOME TABLE: >>> - set IFX_DIRTY_WAIT >>> - open dbaccess , start new transaction (begin work) >>> - lock table in exclusive mode >>> - grant a dummy select over the table (lock sys*) >>> - At other session, kill all users what is accessing this table >>> (identifying with onstat -g opn + onmode -z) >>> - execute the alter over the table >>> - commit >>> If the table have any FK from/to it , I lock and kill users for this >>> tables too. >>> But, at 90% of the situations , after all executed successfully , the >>> system stuck with -710 error. >>> Do not matter if the user close all application and open a new session, >>> the error -710 persist. >>> >>> THE ENVIRONMENT >>> - IFX 11.50 FC9X6 , AIX 6.1 , 4GL 7.32 >>> - Is a centralized 24x7 system what control all plants of the company . >>> - Avg of 1800 sessions concurrent at high peak time and 400 sessions at >>> lower peak time. >>> - User sessions: 70% incoming from 4GL system (most important and where >>> suffer with the problem) >>> - User sessions: 25% incoming from Web/Java system >>> - User sessions: 5% others.... >>> - Database have AUTO_REPREPARE and STMT_CACHE active >>> - The isolation used is 99% 'dirty read' . >>> - The 4GL system work mostly with prepared SQLs where they prepare at >>> beginner of 4GL connection then keep this prepare active during they >>> execution. >>> >>> CONSIDERATIONS >>> - If check with onstat -g dic , always have a refcnt and dirty over the >>> tables involved. This before , during and after the alter . Even during >>> the alter, consider the table in exclusive lock + lock over sys* (grant >>> select). >>> - After the alter, appear new references on onstat -g dic. (with dirty >>> and without dirty) >>> - After finish all alters, if the user got erro -710, close all system , >>> open it again e try again, they stuck at -710 again and again and again >>> and again.... >>> - I try simulate the problem, but with few users, this behave do not >> occur. >>> - Bellow is a stack trace (trapped with onmode -I 710) when the user >>> tryed insert a register over a table where is relationed with the table >>> altered (in this case, added a new index online over columns what not >>> involve the FK) >>> >>> 0x00000001000b76a8 (oninit)afstack >>> >>> 0x00000001000b97bc (oninit)afhandler >>> >>> 0x0000000100130b90 (oninit)check_traperror >>> >>> 0x00000001001faa94 (oninit)sqerr >>> >>> 0x00000001001fa550 (oninit)sqerr1 >>> >>> 0x00000001001f9e54 (oninit)sqnameerr >>> >>> 0x000000010066e4dc (oninit)chkmajvers >>> >>> 0x000000010066eb24 (oninit)openrel >>> >>> 0x00000001006523b4 (oninit)chkparent >>> >>> 0x000000010065b2d0 (oninit)chkrowcons >>> >>> 0x0000000100782790 (oninit)addone >>> >>> 0x0000000100784e10 (oninit)insone_next >>> >>> 0x0000000100769994 (oninit)doinsert >>> >>> 0x000000010041cff8 (oninit)aud_doinsert >>> >>> 0x00000001004259f8 (oninit)excommand >>> >>> 0x000000010044b3bc (oninit)sq_execute >>> >>> 0x000000010022463c (oninit)sqmain >>> >>> 0x000000010037c160 (oninit)listen_verify >>> >>> 0x000000010037a608 (oninit)spawn_thread >>> >>> 0x0000000100dd528c (oninit)startup >>> >>> At my point of view, Informix isn't able to work 100% online if you need >>> to change your scheme (add a simple index using online mode) when you >>> use referential integrity. >>> I believed this is because our method of work (prepare at start of >>> application, and keep this prepared active to reuse it), anyway, should >>> we suffer with this consequences!??? >>> >>> Is there some technique to alter a table what I miss here ? >>> >>> Cesar >>> >>> >>> >>> >> > ******************************************************************************* >>> Forum Note: Use "Reply" to post a response in the discussion forum. >>> >>> >> -- >> Fernando Nunes >> Portugal >> >> http://informix-technology.blogspot.com >> My email works... but I don't check it frequently... >> >> --00235452f64c14f90d04bf0f4faf >> >> >> >> > ******************************************************************************* >> Forum Note: Use "Reply" to post a response in the discussion forum. >> >>