Replication question: docs not clear

[Alan Burlison]

We are trying to automate the startup process for getting replication
running, and the docs are really unclear on if the slave database
*instance* can be running whilst the datafile copy is done, or if it is
just necessary to shut down the *database* that is being replicated on
the slave.

--
Alan Burlison
--

[Alan Burlison]

Alan Burlison wrote:

> We are trying to automate the startup process for getting replication
> running, and the docs are really unclear on if the slave database
> *instance* can be running whilst the datafile copy is done, or if it is
> just necessary to shut down the *database* that is being replicated on
> the slave.

Sorry, that wasn't entirely clear.  The question is which of these two
scenarios is correct:

1. Slave Derby server instance running
2. Database to be replicated is closed on the slave
3. Database data files copied from the master
3. Slave database reopened

or

1. Slave Derby server instance shut down
3. Database data files copied from the master
3. Slave Derby server instance restarted
4. Slave database reopened

--
Alan Burlison
--

[Knut Anders Hatlen]

Alan Burlison <Alan.Burlison@...> writes:

> Alan Burlison wrote:
>
>> We are trying to automate the startup process for getting
>> replication running, and the docs are really unclear on if the slave
>> database *instance* can be running whilst the datafile copy is done,
>> or if it is just necessary to shut down the *database* that is being
>> replicated on the slave.
>
> Sorry, that wasn't entirely clear.  The question is which of these two
> scenarios is correct:
>
> 1. Slave Derby server instance running
> 2. Database to be replicated is closed on the slave
> 3. Database data files copied from the master
> 3. Slave database reopened
>
> or
>
> 1. Slave Derby server instance shut down
> 3. Database data files copied from the master
> 3. Slave Derby server instance restarted
> 4. Slave database reopened

Hi Alan,

Both of those scenarios should work. Whether or not the server is
running while the database is copied doesn't really matter, but the
database itself must not be booted.

By the way, there is a documentation request that describes a slightly
easier procedure to start replication. See
https://issues.apache.org/jira/browse/DERBY-4196. I don't know if it is
relevant in your setup, but I thought I'd mention it in case it could
simplify things.

--
Knut Anders

[Alan Burlison]

Knut Anders Hatlen wrote:

> Both of those scenarios should work. Whether or not the server is
> running while the database is copied doesn't really matter, but the
> database itself must not be booted.

OK, thanks.  Most other databases I've used would get a bit confused if
you started fiddling with the contents of their data file directories
while they were running, it might be worth noting in the docs that this
is OK for Derby.

> By the way, there is a documentation request that describes a slightly
> easier procedure to start replication. See
> https://issues.apache.org/jira/browse/DERBY-4196. I don't know if it is
> relevant in your setup, but I thought I'd mention it in case it could
> simplify things.

Yes, it probably is relevant.  I assume that if you start the master
database with startMaster=true it locks the database until the client is
running, or do you have to manually make sure no clients connect until
after the slave is up and running?

--
Alan Burlison
--

[Knut Anders Hatlen]

Alan Burlison <Alan.Burlison@...> writes:

> Knut Anders Hatlen wrote:
>
>> Both of those scenarios should work. Whether or not the server is
>> running while the database is copied doesn't really matter, but the
>> database itself must not be booted.
>
> OK, thanks.  Most other databases I've used would get a bit confused
> if you started fiddling with the contents of their data file
> directories while they were running, it might be worth noting in the
> docs that this is OK for Derby.
>
>> By the way, there is a documentation request that describes a slightly
>> easier procedure to start replication. See
>> https://issues.apache.org/jira/browse/DERBY-4196. I don't know if it is
>> relevant in your setup, but I thought I'd mention it in case it could
>> simplify things.
>
> Yes, it probably is relevant.  I assume that if you start the master
> database with startMaster=true it locks the database until the client
> is running, or do you have to manually make sure no clients connect
> until after the slave is up and running?

You'd use startMaster=true after you have started the slave, so you
still need to do something (like stopping the server) to prevent clients
from connecting while copying to the slave. I think you'll get a
connection refused exception if you start the master before the slave.

Having a way to start the master and have it block clients until it is
able to reach the slave sounds like a useful improvement, though.

--
Knut Anders