Обсуждение: strange postgresql failure
postgres[666]: [1] LOG: server process (pid 3166) was terminated by signal 11 postgres[666]: [2] LOG: terminating any other active server processes postgres[806]: [591-1] WARNING: Messagefrom PostgreSQL backend: postgres[806]: [591-2] ^IThe Postmaster has informed me that some other backend billy postgres[806]: [591-3] ^Idied abnormally and possibly corrupted shared memory. billy postgres[806]: [591-4] ^II have rolled back the current transaction and am billy postgres[806]: [591-5] ^Igoing to terminate your database system connection and exit. billy postgres[806]: [591-6] ^IPlease reconnect to the database system and repeat your query. After few seconds all backends were disconnected and postgres restarted. It didn't help :-( Even reboot didn't help... Postgres is 7.3.2-2 on Debian Woody. Regards, Tomasz Myrta
Dnia 2003-10-30 12:32, Użytkownik Tomasz Myrta napisał: > After few seconds all backends were disconnected and postgres restarted. > It didn't help :-( Even reboot didn't help... > > Postgres is 7.3.2-2 on Debian Woody. Sorry for my panic. It was my C function which raised segmentation fault. Anyway it would be nice to have some human readable control over C functions... Regards, Tomasz Myrta
Tomasz Myrta wrote: > Dnia 2003-10-30 12:32, Użytkownik Tomasz Myrta napisał: >> After few seconds all backends were disconnected and postgres restarted. >> It didn't help :-( Even reboot didn't help... >> >> Postgres is 7.3.2-2 on Debian Woody. > > Sorry for my panic. It was my C function which raised segmentation fault. > > Anyway it would be nice to have some human readable control over C > functions... There is, sort of ... it's called "core" and if your backend and C function shared object have been compiled with symbol information you can use a debugger to get a stack trace and look at variables. If that's not "human readable" ... well ... I suppose I'm not human. Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com #
Dnia 2003-10-30 18:33, Użytkownik Jan Wieck napisał: > There is, sort of ... it's called "core" and if your backend and C > function shared object have been compiled with symbol information you > can use a debugger to get a stack trace and look at variables. If that's > not "human readable" ... well ... I suppose I'm not human. I know I can load core into gdb... I thought about some kind of message before postgres restarts. I read in syslog that backend terminated with segmentation fault signal and probably corrupted shared memory. It didn't help me to find what really happened. I was wondering if it is possible to know the real reason of restart and leave some message in syslog just before postgres restarts. Regards, Tomasz Myrta