Viewing entries tagged with postgresql. Return to full view.

pginstaller for 8.1

Got together with Dave this monday as I was visiting the UK. Didn't have much time to do anything productive, but we did get some discussion going on what to do about pginstaller for 8.1. Some highlights:

  • User creation has to be moved to the execute sequence. This probably shouldn't be too hard now. It was when we first added it, but things have changed since. This would also mean it could easily be exposed to silent installs, which is a much requested feature.
  • We need to move libpq.dll out of SYSTEM32. The current working model for this is to just remove the ability to move binaries between different directories, and force them into the same directory. Then we just stick libpq.dll there, and the problem is solved. We still need to do some testing on how this will effect the ODBC driver (the new one uses libpq), especially if you have 8.0 installed with libpq.dll in SYSTEM32.
  • Slony will definitly be included, assuming the win32 version is stable and released. But that's the hope.

Of course, we're going to try to tick off anything else that is on the feature tracker on pgfoundry. If anybody wants to help out, now's a great time to start feeding patches :-)

And Dave - nice meeting you! Looking forward to the next time - most likely in Tokyo for the Slony talk!

Slony-win32 service

I have submitted a patch that runs Slony as a service on Win32. It supports multiple parallell engines (if you need to replicate more than one database) if needed.

Still needs a few things polished up (installation/uninstallation, registration of multiple engines etc) but I'd say win32 Slony is progressing very nicely. (Much of these things will probably be available in the installer and pgAdmin as well, but we'll need commandline interfaces as well)

Unicode on win32

As many PostgreSQL users are aware, PostgreSQL 8.0 does not fully support Unicode on win32 due to limitations in the API. Hopefully this will be fixed in 8.1, as we now have two different patches that solves this problem in different ways. The patches were posted a little over a week ago, but due to me sending it to the wrong list it didn't go through because of size. It has now been sent to -patches as well, and is available.

People interested in this are encouraged to apply one or both patches (the native patch is a whole lot easier than the ICU one, as it has much fewer dependencies) and test, test, test! You can grab the patches from the list archive. Let us know your results by a mail to -hackers-win32 or plain -hackers.

Slony on win32 progressing nicely

As of a couple of days ago, the Slony replication engine now runs on Windows.

First of all, this is far from finished work. It's just the initial part that has it up and running and doing simple replication. There are many details still left to iron out. The most noticable ones being support for running as a service and logging to eventlog, but there are others as well.

The changes so far have been smaller than I anticipated. Dave has done a bunch of work on the build system, and the slonik utility has also seen a couple of win32 specific patches not too long ago. But the slon engine is the toughest part, and so far things are looking promising for a patch that's not very disturbing to the rest of the code.

While far from being ready to put in production, if you are looking into deploying Slony on win32 in the future, now is a good time to start playing around with it. Get the latest cvs snapshot, build, and test. And the most important part, give us some feedback on how things are working...

GiST concurrency

Wow. This was pretty fast. On top of the WAL fixes, GiST concurrency fixes are now in, courtesy of Teodor and Oleg (and possibly others? Those are just the ones I know of..) With time to spare before 8.1. Great job guys!

I know at least one of my projects where this feature alone will make me upgrade to 8.1 the day it is released.

Midsummer

Midsummer coming up tomorrow, which means no PostgreSQL hacking for me for a couple of days. (Don't know what midsummer actually means? Visit Sweden for it once, and you'll know..) Once it's over, it's just a couple of days left to the feature freeze...

I got hold of Palle regarding the ICU patch, and it looks like it'll be ready in time for the freeze. Minor things left, and then I'll have to fix up the win32 build issues (which [HTML_REMOVED]can[HTML_REMOVED] be done, as I've done it hackishly before).

The issues of SIGTERM-to-a-backend are back up for discussion. Unsurprisingly, my very hackish idea for a pg_terminate_backend() function based on a flag in shared memory is not going in. Let's just hope that Tom (or someone else, but I'll bet it's going to be Tom eventually) figures out what's bad with the current behaviour (as lock table corruption has been reported) and can get a proper fix in before 8.1.

And I'll not even get into the discussion about the server instrumentation functions at this time...

As for the rest of the open items, I have no idea.. There's a lot of good stuff there, I hope it all gets in.

Happy Midsummer!

Two Phase Commit

It's in. This is one of the big features for 8.1. I'm not sure of any project where I can use it right now myself, but it has been asked for a lot, so I'm sure it will be popular. I'll just have to figure out something to do with it myself.

Looks like we might have a different default database for tools to connect to in 8.1, to get around the "connected to template1, can't create a new database" problems that happen in previous versions.

On top of that, a generic database for tools like pgAdmin to store their config in - wether included by default in an installation or just an agreed-upon shared database name and structure that tools will autocreate.

Lots of activitiy

Lots of activity in the PostgreSQL tree and on the lists lately affecting me directly or indirectly.

Seems we'll have GIST WAL Logging for 8.1. Hopefully we'll also have GIST concurrency. This is great - it'll bring tsearch2 to a whole new level /mha/templates/default/img/emoticons/smile.png alt=":-)" style="display: inline; vertical-align: bottom;" class="emoticon" / Yay!

It also seems the old autovaccuum-in-backend discussion is back up. Let's hope it gets in before the feature freeze this time. It's going to make life a lot easier for a lot of people - especially in the win32 camp where installation was previously a bit of a hassle. If it's done, we'll get it working with the MSI installer out of the box.

Andreas, of pgAdmin fame, is working on getting the backend instrumentation functions in this time - they were rejected for 8.0 because they were submitted too late. While a bit controversial to some people, let's hope they get in this time! Having to rely on contrib modules can be a pain sometimes.

I seem to be all out of pending patches. Both the pending timezone patches are in, and the kerberos work has been in for a while. Now perhaps I have time to look over that signaling patch, and to start digging into Slony.

Also seems the win32 build has been broken for a while. The fact that I didn't notice shows I haven't built on win32 enough lately.

Timezone code finished?

I just finished off the docs part of my latest time zone patch. As Tom appeared at least reasonably positive, I'm hoping for this to go into the codebase shortly.

This patch in itself is fairly simple, and builds on the previous patch that was applied a couple of months back that removed the gobal timezone from the internal API.

PostgreSQL web ftp browser symlink fix

I've just committed a fix to the PostgreSQL website ftp browser to fix handling of symbolic links in the ftp tree.

Previously, symbolic links were only read on creation of the browse pages. This made links to for example /ftp/win32 (which is a symbolic link to /ftp/binary/v8.0.3/win32) fail, since only the /ftp/index.html page was actually updated. Unfortunatly, just this link is present in the release notes for 8.0.3 - and was broken. Not good!

The new code generates Apache rewrite rules that does proper redirection of anything inside a symlinked directory.

Conferences

I speak at and organize conferences around Open Source in general and PostgreSQL in particular.

Upcoming

PGConf.EU 2019
Oct 15-18, 2019
Milan, Italy

Past

Postgres Open 2019
Sep 11-13, 2019
Orlando, FL, USA
Postgres London 2019
Jul 2-3, 2019
London, UK
PGCon 2019
May 27-31, 2019
Ottawa, Canada
PGDay.IT 2019
May 17, 2019
Bologna, Italy
PGConf.DE 2019
May 10, 2019
Leipzig, Germany
More past conferences