[Bucardo-general] Bucardo question

Женя Деревянкин jekader at gmail.com
Tue Dec 8 22:51:55 UTC 2009


Hi, Greg.

Thanks for the info.

My question is in making any of the two remaining servers syncronised.
So, if I have three servers, A, B and C, they'e sync'ed the following way:
A => B
A => C

All of them have read-write access, and produce database changes.

So, if A goes down, I must start sync'ing B => С.
When A goes back online, the db contents are manually cleared,
pulled from B, and A becomes a "slave", so we have:
B => С
B => A

Until B fails, and the system returns to the initial state.

Of cource, that's the worst-case scenario. The chances that specifically the
"master" database fails are pretty low, but I can't exclude such a possibility.

So I wanted to know if it's safe to switch "masters" in a way I've described.
As I've understood, the delta tables just accumulate changes since last sync,
so it should work, but I could be missing something.

--
Best regards,
Eugen

2009/12/8 Greg Sabino Mullane <greg at endpoint.com>:
>> So, my goal is to make several "clone" databases (3 or more), which
>> must be kept in sync even after one of them fails. After recovery the
>> database can be put in operation manually.
>>
>> But if the server that fails is the one bucardo is running on? What
>> must happen then? I mean not just that postgres dies, but the whole
>> server. A new instance of bucaurdo should start running on one of the
>> remaining servers, keeping the data synced? Won't that interfere into
>> the database structure?
>
> You can put Bucardo anywhere, and have backup copies on other servers
> ready to go. While you could even use Bucardo to replicate the bucardo
> database :), it's small enough that a pg_dump will suffice.
>
> It's unclear what your scenario is as far as failing slaves, but Bucardo
> should be able to handle that. In the future it will be automatic, but
> generally, if a slave goes down, you set that database as "inactive"
> inside Bucardo and restart it. When it comes back up, just set the db to
> "active" and restart Bucardo again.
>
> --
> Greg Sabino Mullane greg at endpoint.com
> End Point Corporation
> PGP Key: 0x14964AC8


More information about the Bucardo-general mailing list