Re: Hot standby from Debian to Windows

On 3/29/20 12:52 PM, Andrus wrote:
> Hi!
> 
>> Since you are moving between different OSes you will need to use some 
>> form of logical replication as binary replication will not work.
> 
> I can use Hyper-V or something other to run Debian with Windows.
> 
> This hopefully will also allow to bypass Windows 20 connection limit so 
> that more than 20 users can connect.
> 
>> Given  that you are Postgres 12 you could use the builtin logical 
>> replication:
>> https://www.postgresql.org/docs/12/logical-replication.html
> 
> I see possibilities:
> 
> 1. Use Hyper-V to run Debian container and Postgres binary replication
> 2. Use Hyper-V to run Debian container and Postgres logical replication
> 3. Use Windows and Postgres logical replication.
> 4. Use Hyper-V to run Debian container and backup utility  for whole 
> disk block level backup
> 
>> How that needs to managed is going to need more information. As a start:
>> 1) Downtime allowed?
> 
> Yes. If main server stops, I can tell users to enter backup server 
> address instead of main server.
> 
>> 2) All databases to be replicated or just some?
> 
> I can create separate cluster so that all databases and users will be 
> replicated.
> There are 30 databases with total size 70 GB. Size will increase slowly 
> every day when new data is entered.
> There are some test and demo databases whose replcation is not really 
> required but those can also replicated if this
> 
>> 3) Permissible lag between servers?
> 
> Currently backups are created every night and restored in new server.
> Backup of 67GB data takes 1 hour, transfer 1 hour, restore and analyze 
> to new server 4 hours. Total 6 hours. So current lag in 6 .. 24 hours.
> 
> Goal is to decrease this lag.
> 
>> 4) How are you going to deal with the down server and how do you plan 
>> on bringing it up again?
> 
> VPS hosting company will bring it up again. I will then manually 
> synchronize two clusters when users continue to enter data, this is not 
> time critical.

Would it not be easier to just set up another Debian server, run binary 
replication and put them behind something like pgpool?

> 
> Andrus.


-- 
Adrian Klaver
adrian.klaver@aklaver.com