[Bucardo-general] Automatic Cascading Replication

[Greg Sabino Mullane]

On Tue, Dec 18, 2012 at 08:39:21PM -0800, David E. Wheeler wrote:
> Because if A goes down, frobishers won’t be updated until it comes back up. 
> I want frobishers to always be as up-to-date as whatever is currently 
> running in the MM “cluster.”

Frobishers (let's call it C, ok? :) will not be up to date if A goes down, 
if the second sync is:

table=customers dbs=A:source B:source C:target

Bucardo cannot push out to C unless A and B are both available. What you 
would need is three syncs:

1) MM between lots of tables A:source B:source
2) One-way customer table only A:source C:target
3) One-way customer table only B:source C:target

And of course customer table should have makedelta:A,B

>> We discussed on #bucardo having triggers exit if session_replication_role 
>> is set to "local", and to set it to that for bulk inserts.

Well, even if one were to do this (and I am not convinced), we would 
need to use seomthing other than session_replication_role. Because 
setting it to "local" means that all *other* rules and triggers 
would fire, something definitely not wanted.

>> Yeah, but that still means firing that trigger for every single row 
>> on the bulk insert. Not so bulky anymore. :)
> 
> I don’t know what the actual overhead on that is, do you? I mean, 
> compared to the I/O required when a COPY happens, is it significant?

I don't know the numbers, no.

-- 
Greg Sabino Mullane greg at endpoint.com
End Point Corporation
PGP Key: 0x14964AC8
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 163 bytes
Desc: not available
URL: <https://mail.endcrypt.com/pipermail/bucardo-general/attachments/20121219/d70b990b/attachment.sig>