Thread: bugs or no in PostgreSQL 7.1 ?

bugs or no in PostgreSQL 7.1 ?

From
"Andrey Mikuloff"
Date:
After compiling and porting database from 7.0.3 to 7.1 perfomance was cool,
but later, when many clients connected to our server and process
transactions (creating documents e.t.c.) PostgreSQL 7.1 say's "ERROR:
Cannot insert a duplicate key into unique index pay_docs_list_pkey", but all
indexes and sequences was OK, and after vacuum, reindexing base and
increasing last number of sequence (defined as "serial") errors was
repeated.
And also, time to time, Postgres die with logs:
Server process (pid ...) exited with status 139 at ... Terminating any
active server processes...
        The Postmaster has informed me that some other backend  died
abnormally and possibly corrupted shared memory.         I have rolled back
the current transaction and am going to terminate your database system
connection and exit.  Please reconnect to the database system and repeat
your query.

This error usualy after vacuum small tables on the fly

And in final PostgreSQL 7.1 kill our databse, i.e. can't restart on they
database with error - can't restore pg_xlog records...

More: what is error on start PostgreSQL 7.1: [: 11800: unary operator
expected
it's sometime displaing, but not always ?

Postgresql was compiled with
params  --enable-locale --enable-recode --with-maxbackends=1024, data was in
koi8 with recode by charset.conf. Also php4.0.4pl1

What occurs???????????


May be some bugs with die or no answer of backend becouse used 'recode' ?
I think that 'recode' more quickly than 'multibyte', or no ?

Sorry my bad english, can answer in russian or english

Andrey Mikuloff









Re: bugs or no in PostgreSQL 7.1 ?

From
Tom Lane
Date:
"Andrey Mikuloff" <mikuloff@kazna.ru> writes:
> More: what is error on start PostgreSQL 7.1: [: 11800: unary operator
> expected
> it's sometime displaing, but not always ?

I think that's a known bug in some system-specific startup script
(sloppy quoting, or something like that).  Should be fixed in 7.1.1.

> May be some bugs with die or no answer of backend becouse used 'recode' ?

Possibly; the recode support is not as well tested as some other parts
of the system.  But it's premature to jump to any conclusions with so
little data.  Can you provide backtraces from the core dump files?
Also, if it won't restart, try setting WAL_DEBUG=1 to get a more verbose
indication of what's failing in the restart.

            regards, tom lane