Re: [HACKERS] VACUUM ANALYZE problem on linux

Oleg Broytmann <phd(at)sun(dot)med(dot)ru> writes:
>> I can test on libc5 if you still see trouble after you have verified
>> Jan's fixes for your memory exhaution problem.

> I've downloaded latest snapshot (9 Feb) and reproduced the problem with
> VACUUM ANALYZE on Debian 2.0 (glibc2).

I am not able to reproduce the problem on HPUX, using either current
sources or 6.4.2. Looks like it must be platform specific.

Could you build the backend with -g and send a gdb backtrace from the
corefile produced when the crash occurs?

regards, tom lane

On Thu, 11 Feb 1999, Tom Lane wrote:
> I am not able to reproduce the problem on HPUX, using either current
> sources or 6.4.2. Looks like it must be platform specific.

Of course it is platform-specific. I reported the problem on
glibc2-based linucies, but the same database works fine (and allows VACUUM
ANALYZE) on sparc-solaris.
Don't know about libc5 linux - I have no one in hand.

> Could you build the backend with -g and send a gdb backtrace from the
> corefile produced when the crash occurs?

I'll do it this Saturday.

Oleg.
----
Oleg Broytmann National Research Surgery Centre http://sun.med.ru/~phd/
Programmers don't die, they just GOSUB without RETURN.

Hi!

On Thu, 11 Feb 1999, Tom Lane wrote:
> Could you build the backend with -g and send a gdb backtrace from the
> corefile produced when the crash occurs?

Problem compiling with -g:

make[3]: Entering directory
`/usr/local/src/PostgreSQL/postgresql-6.4.2/src/interfaces/ecpg/preproc'
gcc -I../../../include -I../../../backend -g -Wall -Wmissing-prototypes
-I../include -DMAJOR_VERSION=2 -DMINOR_VERSION=4 -DPATCHLEVEL=4
-DINCLUDE_PATH=\"/usr/local/stow/pgsql-debug/include\"
-c y.tab.c -o y.tab.o
preproc.y:2389: parse error at end of input
preproc.y:20: warning: `struct_level' defined but not used
preproc.y:22: warning: `QueryIsRule' defined but not used
preproc.y:23: warning: `actual_type' defined but not used
preproc.y:24: warning: `actual_storage' defined but not used
preproc.y:219: warning: `remove_variables' defined but not used
preproc.y:254: warning: `reset_variables' defined but not used
preproc.y:263: warning: `add_variable' defined but not used
preproc.y:332: warning: `make1_str' defined but not used
preproc.y:341: warning: `make2_str' defined but not used
preproc.y:353: warning: `cat2_str' defined but not used
preproc.y:366: warning: `make3_str' defined but not used
preproc.y:380: warning: `cat3_str' defined but not used
preproc.y:396: warning: `make4_str' defined but not used
preproc.y:412: warning: `cat4_str' defined but not used
preproc.y:431: warning: `make5_str' defined but not used
preproc.y:449: warning: `cat5_str' defined but not used
preproc.y:471: warning: `make_name' defined but not used
preproc.y:481: warning: `output_statement' defined but not used
make[3]: *** [y.tab.o] Error 1
make[3]: Leaving directory
`/usr/local/src/PostgreSQL/postgresql-6.4.2/src/interfaces/ecpg/preproc'

Oleg.
----
Oleg Broytmann National Research Surgery Centre http://sun.med.ru/~phd/
Programmers don't die, they just GOSUB without RETURN.

> > Could you build the backend with -g and send a gdb backtrace from
> > the corefile produced when the crash occurs?
> Problem compiling with -g:

I'd be suprised if "-g" would do that to you. Are you sure the input
file is well-formed?

- Tom

Hi!

On Sat, 13 Feb 1999, Thomas G. Lockhart wrote:
> > > Could you build the backend with -g and send a gdb backtrace from
> > > the corefile produced when the crash occurs?
> > Problem compiling with -g:
>
> I'd be suprised if "-g" would do that to you. Are you sure the input
> file is well-formed?

May be not enough memory or disk space. In the beginnig of next week
I'll have new linux box, so I'll retry.

Oleg.
----
Oleg Broytmann National Research Surgery Centre http://sun.med.ru/~phd/
Programmers don't die, they just GOSUB without RETURN.

Hello!

Sorry for making this late :(

On Thu, 11 Feb 1999, Tom Lane wrote:
> Could you build the backend with -g and send a gdb backtrace from the
> corefile produced when the crash occurs?

I have compiled with -g, but postgres didn't produce core. Do I need
something special on startup to generate core on crash?

Oleg.
----
Oleg Broytmann http://members.xoom.com/phd2/ phd2(at)earthling(dot)net
Programmers don't die, they just GOSUB without RETURN.

Hello!

On Thu, 11 Feb 1999, Tom Lane wrote:
> Could you build the backend with -g and send a gdb backtrace from the
> corefile produced when the crash occurs?

I have a problem getting core - postgres didn't produces core. Recently
I got a suggestion (from Vadim) to attach gdb to a running process and
debug it this way. Ok, done.
To remind of the problem - I have a problem running VACUUM ANALYZE on a
glibc2 linux (Debian 2.0). On solaris it is Ok (and I got a report it is Ok
on HP-UX).
Here is the traceback. The problem is in strcoll, don't understand why.

Program received signal SIGSEGV, Segmentation fault.
0x40119587 in strcoll ()
(gdb) where
#0 0x40119587 in strcoll ()
#1 0x816cadd in varstr_cmp (arg1=0x4020fccc " ", len1=0,
arg2=0x8268604 " ", len2=0) at varlena.c:511
#2 0x816b31d in bpcharlt (arg1=0x4020fcc8 "\n", arg2=0x8268600 "\n")
at varchar.c:504
#3 0x80a1bb7 in vc_attrstats (onerel=0x8264378, vacrelstats=0x8262a10,
tuple=0x4020fc90) at vacuum.c:1630
#4 0x809ffec in vc_scanheap (vacrelstats=0x8262a10, onerel=0x8264378,
vacuum_pages=0xbfffcf84, fraged_pages=0xbfffcf78) at vacuum.c:806
#5 0x809f773 in vc_vacone (relid=32600, analyze=1 '\001', va_cols=0x0)
at vacuum.c:504
#6 0x809ed84 in vc_vacuum (VacRelP=0x0, analyze=1 '\001', va_cols=0x0)
at vacuum.c:257
#7 0x809ec3e in vacuum (vacrel=0x0, verbose=0 '\000', analyze=1 '\001',
va_spec=0x0) at vacuum.c:160
#8 0x8138c43 in ProcessUtility (parsetree=0x82464e0, dest=Remote)
at utility.c:644
#9 0x8135388 in pg_exec_query_dest (query_string=0xbfffd0d4 "vacuum
analyze;",
dest=Remote, aclOverride=0 '\000') at postgres.c:758
#10 0x8135264 in pg_exec_query (query_string=0xbfffd0d4 "vacuum analyze;")
at postgres.c:699
#11 0x813677e in PostgresMain (argc=6, argv=0xbffff15c, real_argc=8,
real_argv=0xbffffafc) at postgres.c:1645
#12 0x8111561 in DoBackend (port=0x82110f0) at postmaster.c:1541
#13 0x8110f35 in BackendStartup (port=0x82110f0) at postmaster.c:1312
#14 0x811012f in ServerLoop () at postmaster.c:757
#15 0x810fb5e in PostmasterMain (argc=8, argv=0xbffffafc) at
postmaster.c:563
#16 0x80c8c32 in main (argc=8, argv=0xbffffafc) at main.c:93

>
> regards, tom lane
>

Oleg.
----
Oleg Broytmann http://members.xoom.com/phd2/ phd2(at)earthling(dot)net
Programmers don't die, they just GOSUB without RETURN.