> What do you find difficult about the current method? That's got to be
> the first discussion point.
The main problem I have is the complexity of setup.
It involves a lot of additional scripting which you have to get it right
to be actually reliable. The documentation is giving a rough idea on how
to do it, but it is quite some work to make it work, and you can't
really tell that is reliable...
Another issue is that unless you got the archive_command right in the
master server from the beginning, you will have to restart the server
once you decide to build your standby... the archive_command is a
start-up time parameter. This could be of course alleviated by always
using a stub script as archive command, and let it do nothing if you
don't have a standby, and then modify it to start archiving to the right
place once you start building one, or if you want to move it to another
machine. But this is also not documented, and you have to figure it out
for yourself.
And responding to Tom's other post regarding a real standby mode, where
you could stop the standby and then later resume it still in standby
mode: I would actually have a good use for it tonight :-)
We will migrate our application to a new version, which involves some
changes in the data base. Now it would be nice to stop the standby
BEFORE doing these changes, and if the migration fails for some reason,
start up the standby and use it with our old application version. But if
the migration succeeds, I want to start up the standby still as standby,
and make it resume standby operation... rebuilding it will take half day
at least.
So a standby which can be isolated for a while would actually be useful.
OK, now that I'm thinking a bit more about this, I could achieve this by
fiddling with the restore_command so it stops delivering the logs for a
while. But again it is not straightforward.
The whole point of this is that starting up a standby should be as
simple as pointing the standby machine to the primary server, without
shell scripting gimmicks (which are OS specific and therefore hard to
document in a generic way), without the need of fiddling with the
primary's configuration (see archive command), without the need to
restart the primary if the archive command was not right in the first
place. And to make it easy to start up one more standby if needed, or
isolate it for a while when doing some risky work on the primary.
It's about user friendliness and flexibility. It's not that it can't do
the work right now, but it's really hard to do it...
Cheers,
Csaba.