PostgreSQL Replication Server?

On Fri, 1 Jun 2001, Gerald Gutierrez wrote:

The database article at Zend mentioned a replication server for
PostgreSQL., available at www.pgsql.com. The press release says the code
itself has been put into the CVS tree, but I can't find any documentation
on it anywhere.

You won't find any documentation, for it seems that none has been
written. I was bitten by this fact and had to use an ugly hack to make
replication work... without erserver (basically dump and restore at the
other side). It works as long as the database is small...

Now, erserver seems to work, but it needs a bit hacking around that I
hadn't done yet. Maybe when I get it working I'll see to writing
something. In the mean time, source code is the only thing that can help
you.

The code is in contrib/rserv.

--
Alvaro Herrera (<alvherre[@]atentus.com>)

On Wed, 6 Jun 2001, Alvaro Herrera wrote:

On Fri, 1 Jun 2001, Gerald Gutierrez wrote:

Now, erserver seems to work, but it needs a bit hacking around that I
hadn't done yet. Maybe when I get it working I'll see to writing
something. In the mean time, source code is the only thing that can help
you.

I forgot to explain: erserver works well "out of the box" for one
master / one slave setup. If you need more than one slave, you need some
hacking around. If you need more than one master, I doubt it'll work in
its current form.

--
Alvaro Herrera (<alvherre[@]atentus.com>)

Or try:

http://pgreplicator.sourceforge.net/

Haven't used it myself yet, but it looks pretty good...

Now, erserver seems to work, but it needs a bit hacking around

that I

hadn't done yet. Maybe when I get it working I'll see to writing
something. In the mean time, source code is the only thing that

can help

you.

I forgot to explain: erserver works well "out of the box" for one
master / one slave setup. If you need more than one slave, you need

some

hacking around. If you need more than one master, I doubt it'll work

in

its current form.

--
Alvaro Herrera (<alvherre[@]atentus.com>)

---------------------------(end of

broadcast)---------------------------

On Thu, Jun 07, 2001 at 09:06:06AM +0100, Gordan Bobic wrote:

Or try:

http://pgreplicator.sourceforge.net/

Haven't used it myself yet, but it looks pretty good...

I tried to build it , but it depends on an extention to tcl (tcl-dp)
that I couldn't get working. Admittedly, I gave up pretty easily.
But just a warning that you need something which might not be
available everywhere, whereas the rserv code depends on Perl.

-- 
----
Andrew Sullivan                               87 Mowat Avenue 
Liberty RMS                           Toronto, Ontario Canada
<andrew@libertyrms.com>                               M6K 3E3
                                         +1 416 646 3306 x110

The database article at Zend mentioned a replication server for
PostgreSQL., available at www.pgsql.com. The press release says the code
itself has been put into the CVS tree, but I can't find any documentation
on it anywhere.

The article at Zend also stated that PostgreSQL could handle 1000
transactions a second with 100 users. This is funny, since my postgresql
server (7.1.2) running on a PII 533 with SCSI and 256 mb RAM (HP Server)
can't even handle 1000 select statements per second. Not even close.

Daniel Akerud

The database article at Zend mentioned a replication server for
PostgreSQL., available at www.pgsql.com. The press release says the

code

itself has been put into the CVS tree, but I can't find any

documentation

on it anywhere.

The article at Zend also stated that PostgreSQL could handle 1000
transactions a second with 100 users. This is funny, since my postgresql
server (7.1.2) running on a PII 533 with SCSI and 256 mb RAM (HP Server)
can't even handle 1000 select statements per second. Not even close.

"transactions per second" does not equate to "real-world queries per
second". : )

When we were sizing out a machine, a Compaq rep came in and asked how
many queries we did in a day, and he told us that a single CPU alpha could
handle about ten times that many "transactions per second". The machine
we have now is a lot faster than a single Alpha, and is getting
overloaded.

steve

On Thu, 7 Jun 2001 zilch@home.se wrote:

The database article at Zend mentioned a replication server for
PostgreSQL., available at www.pgsql.com. The press release says the code
itself has been put into the CVS tree, but I can't find any documentation
on it anywhere.

The article at Zend also stated that PostgreSQL could handle 1000
transactions a second with 100 users. This is funny, since my postgresql
server (7.1.2) running on a PII 533 with SCSI and 256 mb RAM (HP Server)
can't even handle 1000 select statements per second. Not even close.

Small machine, sorry ... move to a Dual-PIII with a gig of RAM and
multiple drives striped ... bump the shared memory to use 512Meg of RAM so
that alot of caching is being done in memory ... then you have a serious
server :)

we did some testing where a client was seeing 4k+ TPS *shrug*

The Hermit Hacker wrote:

On Thu, 7 Jun 2001 zilch@home.se wrote:

The database article at Zend mentioned a replication server for
PostgreSQL., available at www.pgsql.com. The press release says the code
itself has been put into the CVS tree, but I can't find any documentation
on it anywhere.

The article at Zend also stated that PostgreSQL could handle 1000
transactions a second with 100 users. This is funny, since my postgresql
server (7.1.2) running on a PII 533 with SCSI and 256 mb RAM (HP Server)
can't even handle 1000 select statements per second. Not even close.

Small machine, sorry ... move to a Dual-PIII with a gig of RAM and
multiple drives striped ... bump the shared memory to use 512Meg of RAM so
that alot of caching is being done in memory ... then you have a serious
server :)

we did some testing where a client was seeing 4k+ TPS *shrug*

So, is one select statement equal to one of these transactions? Seems to me
that talking about transactions per second is like talking about instructions
per second on a complex instruction-set CPU - depending what you use
for an instruction, your answer could vary significantly.

we did some testing where a client was seeing 4k+ TPS *shrug*

So, is one select statement equal to one of these transactions? Seems to

me

that talking about transactions per second is like talking about

instructions

per second on a complex instruction-set CPU - depending what you use
for an instruction, your answer could vary significantly.

Just reading this discussion it sounds to me like empty transactions (ie. a
bunch of BEGIN; COMMIT;). Which is a good measure of transaction overhead,
but not necessarily real world performance.

Greg