Обсуждение: 7.2b3 pg_dump, general 7.2b3 comments
Is pg_dump supposed to work against prior version databases? I was moving data from an old server (running 7.2devel from several months ago) to a new server (running 7.2b3), using pg_dump on the *new* server, and got the following message: # pg_dump -h 172.16.1.84 -U postgres -t el_names -i lt_lcat > el_names.dmp Password: <types in password> pg_dump: query to obtain list of tables failed: ERROR: Attribute 'relhasoids' not found Not that I need to do this, but it was convenient and I thought it was supposed to work. Is this a problem, or am I trying to do something unsupported? On a side note, I was *very* happy when I was able to load a table with ~40 million rows about 30 minutes (compared to about a day+ on the old hardware, Red Hat 6.2, and early 7.2devel Postgres). And that was without changing the default postgresql.conf, because I forgot to do it before I started the copy ;-). Memory usage according to 'top' never exceeded about 5 MB. Also worth noting, I installed successfully from an RPM which I built from the source RPM. Thank you to *everyone* involved in getting PostgreSQL to where it is today! It is truly an awesome product. -- Joe
Joe Conway <joseph.conway@home.com> writes:
> Is pg_dump supposed to work against prior version databases?
Since 7.1.1 we've tried to make it do so.
> # pg_dump -h 172.16.1.84 -U postgres -t el_names -i lt_lcat > el_names.dmp
> Password: <types in password>
> pg_dump: query to obtain list of tables failed: ERROR: Attribute
> 'relhasoids' not found
<scratches head> Odd. There's only one query in pg_dump that touches
relhasoids, and it's set up to only be used when "remoteVersion >= 70200".
It seems to work here, too: I can dump from a 7.1 or even 7.0 server
with current pg_dump. Would you burrow in there and see what's screwing
up the version-check code?
BTW, what happens if you leave off -i? It shouldn't be necessary.
regards, tom lane
Joe Conway <joseph.conway@home.com> writes:
> Is pg_dump supposed to work against prior version databases? I was
> moving data from an old server (running 7.2devel from several months
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> ago) to a new server (running 7.2b3), using pg_dump on the *new* server,
> and got the following message:
> # pg_dump -h 172.16.1.84 -U postgres -t el_names -i lt_lcat > el_names.dmp
> Password: <types in password>
> pg_dump: query to obtain list of tables failed: ERROR: Attribute
> 'relhasoids' not found
Oh, never mind: I see it. pg_dump's test is on whether the server calls
itself 7.2 or not. You've evidently got a copy from back before the
changes to make OIDs optional. The versioning code is not set up to
deal with intermediate development versions, so it gets it wrong about
what query to use. You'll need to use the pg_dump of the same vintage
as the 7.2devel server.
You're a brave man to be putting production data on CVS-tip servers.
I wouldn't recommend it ;-)
regards, tom lane
Tom Lane wrote: > Joe Conway <joseph.conway@home.com> writes: > >>Is pg_dump supposed to work against prior version databases? I was >>moving data from an old server (running 7.2devel from several months >> > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > >>ago) to a new server (running 7.2b3), using pg_dump on the *new* server, >>and got the following message: >> > >># pg_dump -h 172.16.1.84 -U postgres -t el_names -i lt_lcat > el_names.dmp >>Password: <types in password> >>pg_dump: query to obtain list of tables failed: ERROR: Attribute >>'relhasoids' not found >> > > Oh, never mind: I see it. pg_dump's test is on whether the server calls > itself 7.2 or not. You've evidently got a copy from back before the > changes to make OIDs optional. The versioning code is not set up to > deal with intermediate development versions, so it gets it wrong about > what query to use. You'll need to use the pg_dump of the same vintage > as the 7.2devel server. > > You're a brave man to be putting production data on CVS-tip servers. > I wouldn't recommend it ;-) > > regards, tom lane > Thanks, but I'm not that brave. This has been my proof-of-concept server for the past 6 or so months. The new server is intended to become a production, relatively high volume data collection server on our factory floor. I've finally been successful at selling our upper management on using PostgreSQL in lieu of commercial databases for this purpose (we already have multiple instances of brand O and brand M). In fact, given the current economy, they're enthusiastic about it ;-) In any case, I've already moved the data, just wanted to report the possible issue. It sounds like it shouldn't affect any but the highly adventurous! BTW, after your first reply, I started to load the new pg_dump into gdb and discovered it had no debug symbols (recall I installed from RPM). Is there a way to install the RPM with additional configure options without rebuilding it? Is there any significant downside (performance or otherwise) to having --enable-debug on a production server? -- Joe
Joe Conway <joseph.conway@home.com> writes:
> BTW, after your first reply, I started to load the new pg_dump into gdb
> and discovered it had no debug symbols (recall I installed from RPM). Is
> there a way to install the RPM with additional configure options without
> rebuilding it?
Don't know; certainly you'd have to recompile, but I dunno if you have
to modify the source RPM or not. Lamar?
> Is there any significant downside (performance or
> otherwise) to having --enable-debug on a production server?
If you're compiling with gcc then I believe the only cost is the disk
footprint of the debug info. On some other compilers, --enable-debug
disables most compiler optimizations, which can mean a significant
speed penalty. We currently have the following in the installation
guide:
--enable-debug
Compiles all programs and libraries with debugging symbols. This means that you can run the programs
througha debugger to analyze problems. This enlarges the size of the installed executables considerably,
andon non-GCC compilers it usually also disables compiler optimization, causing slowdowns. However,
havingthe symbols available is extremely helpful for dealing with any problems that may arise. Currently,
thisoption is considered of marginal value for production installations, but you should have it on if you
aredoing development work or running a beta version.
--enable-cassert
Enables assertion checks in the server, which test for many "can't happen" conditions. This is
invaluablefor code development purposes, but the tests slow things down a little. Also, having the tests
turnedon won't necessarily enhance the stability of your server! The assertion checks are not categorized
forseverity, and so what might be a relatively harmless bug will still lead to server restarts if it triggers
an assertion failure. Currently, this option is not recommended for production use, but you should have it
onfor development work or when running a beta version.
Perhaps "marginal value" is too lukewarm an assessment, at least for
gcc users. Comments?
regards, tom lane
Tom Lane wrote: > Perhaps "marginal value" is too lukewarm an assessment, at least for > gcc users. Comments? > Well, if I have problems on my production machine, and I post a question here, the first thing I'll be asked for is a backtrace, preferably with symbols, right ;-) So ISTM, that if there is no penalty except (a relatively trivial amount of) disk space with gcc, then that ought to be the preferred configuration for gcc. The "value" for a production installation is significant if it helps get a problem fixed faster. -- Joe
On Thursday 29 November 2001 01:27 am, Tom Lane wrote: > Joe Conway <joseph.conway@home.com> writes: > > BTW, after your first reply, I started to load the new pg_dump into gdb > > and discovered it had no debug symbols (recall I installed from RPM). Is > > there a way to install the RPM with additional configure options without > > rebuilding it? > Don't know; certainly you'd have to recompile, but I dunno if you have > to modify the source RPM or not. Lamar? Hmmm. You know, it would be a good idea, IMHO, to enable debugging symbols in the beta RPMs anyway. So, I will do that for the next beta (or release candidate -- although, with that report from Jan, I wonder what the next release will be). In the meantime, Joe, if you can rebuild the RPM from source, here's what to do: 1.) rpm -i the source RPM. 2.) If you've never built from a source RPM before, see /usr/share/docs/postgresql-7.2b3/README.rpm-dist for some more information. 3.) Edit the spec file (on Red Hat, that would be in /usr/src/redhat/SPECS, name of 'postgresql.spec'), adding the following line near the top: %define __os_install_post /usr/lib/rpm/brp-compress 4.) Add the configure option. The rpm build process by default runs strip on the binaries..... 5.) rpm -ba postgresql.spec, wait a few minutes for the build, and pick up your RPMs in /usr/src/redhat/RPMS/i386. Or you can just wait until I upload a set with that line in it..... :-) -- Lamar Owen WGCR Internet Radio 1 Peter 4:11
On Thursday 29 November 2001 07:59 pm, Lamar Owen wrote: > Or you can just wait until I upload a set with that line in it..... :-) Said set is now being uploaded to ftp.postgresql.org. Look for the 0.3PGDG sbinary et in /pub/binary/beta/RPMS/redhat-7.2, and the source in ......./SRPMS My my, --enable-debug and disabling stripping in the build sure does inflate the package size :-O For reference: The entire binary set for 7.2b3-0.2PGDG weighs in at 7,036KB. The debug-enabled 7.2b3-0.3PGDG set weighs in at 11,240KB. The biggest increase is in the postgresql-contrib package, which increases from 1,192KB to 2,940KB. This build also has --enable-cassert. -- Lamar Owen WGCR Internet Radio 1 Peter 4:11
Lamar Owen wrote: > Hmmm. You know, it would be a good idea, IMHO, to enable debugging symbols in > the beta RPMs anyway. So, I will do that for the next beta (or release > candidate -- although, with that report from Jan, I wonder what the next > release will be). > > In the meantime, Joe, if you can rebuild the RPM from source, here's what to > do: > 1.) rpm -i the source RPM. > 2.) If you've never built from a source RPM before, see > /usr/share/docs/postgresql-7.2b3/README.rpm-dist for some more information. > 3.) Edit the spec file (on Red Hat, that would be in /usr/src/redhat/SPECS, > name of 'postgresql.spec'), adding the following line near the top: > %define __os_install_post /usr/lib/rpm/brp-compress > 4.) Add the configure option. The rpm build process by default runs strip on > the binaries..... > 5.) rpm -ba postgresql.spec, wait a few minutes for the build, and pick up > your RPMs in /usr/src/redhat/RPMS/i386. > > Or you can just wait until I upload a set with that line in it..... :-) > Thanks for the instructions, Lamar -- that helps quite a bit. I'll give this a try next week. Joe
On Friday 30 November 2001 08:50 pm, Joe Conway wrote: > Lamar Owen wrote: > > Or you can just wait until I upload a set with that line in it..... :-) > Thanks for the instructions, Lamar -- that helps quite a bit. I'll give > this a try next week. While you're certainly welcome to do the work yourself, you are also equally welcome to download the latest RPMset, which has debugging symbols enabled, AFAIK. -- Lamar Owen WGCR Internet Radio 1 Peter 4:11
Lamar Owen wrote: >>Lamar Owen wrote: > While you're certainly welcome to do the work yourself, you are also equally > welcome to download the latest RPMset, which has debugging symbols enabled, > AFAIK. > Thanks, I saw that, but I'd like to know how to do these kinds of adjustments to RPMs anyway. I normally prefer to install from source tarballs, or even CVS, but for production servers our network admins get a warm fuzzy from RPMs. It gives them more confidence they can reproduce the server setup if they need to after I hit the Lotto and retire ;-) Joe