[Bucardo-general] Failover, production env use
Greg Sabino Mullane
greg at endpoint.com
Wed Nov 5 18:26:34 UTC 2008
> My other quation belongs to the main (and only?) developer, Greg Sabino
> Mullane. Ive seen some pdf file on the www, what were slide's to a pg
> conn. On that i saw something like "Not for failover". Is that true? (Or
> did u mean it: its not a cluster-solution, where one node fails clients
> ll be served via oder nodes? Wich is clear)
Not the only developer, but the most active and visible one at the
moment. :) By "no failover" I mean that it cannot be easily switched so
that another node becomes a new master, like Slony does. Something like
that is possible with Bucardo, but not in anything resembling an automatic
system at the moment.
> The environment in i would use bucardo is: geographically distributed
> nodes, apps based only on local nodes only, rapdid synchronized
> "goats"(in a couple of seconds), and the main req: on resume from
> connection faliure's (wich is tipically few hours, but it can be days
> too) fast recovery. (the busines logic layer upon the ddb is still
> working until unsychronised) The size of the distributed db is so around
> 1-2GB, but the "to-sync-data-per-day" is very low (few 10 megabyte worst
> case). Would u recommend bucardo for that purpose?
Yes, Bucardo is particularly good at catching up from long
disconnections, as it merely grabs a list of which rows have changed, and
doesn't care how many transactions actually occurred to get them that way.
Greg Sabino Mullane greg at endpoint.com
End Point Corporation 610-983-9073
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 163 bytes
Desc: not available
Url : https://mail.endcrypt.com/pipermail/bucardo-general/attachments/20081105/c3519ff3/attachment.bin
More information about the Bucardo-general