Andres, nice job on the writeup.<br /><br /> I think one aspect you are missing is that there must be some way for the
multi-mastersto <br /> re-stabilize their data sets and quantify any data loss. You cannot do this without<br /> some
replicationintelligence in each row of each table so that no matter how disastrous<br /> the hardware/internet failure
inthe cloud, the system can HEAL itself and keep going, no human beings involved.<br /><br /> I am laying down a
standarddesign pattern of columns for each row:<br /><br /> MKEY - Primary key guaranteed unique across ALL nodes in
theCLOUD with NODE information IN THE KEY. (A876543 vs B876543 or whatever)(network link UP or DOWN)<br /> CSTP -
createtime stamp on unix time stamp<br /> USTP - last update time stamp based on unix time stamp<br /> UNODE - Node
thatupdated this record<br /><br /> Many applications already need the above information, might as well standardize it
soexternal replication logic processing can self heal.<br /><br /> Postgresql tables have optional 32 bit int OIDs, you
maywant consider having a replication version of the ROID, replication object ID and then externalize the primary<br />
keygeneration into a loadable UDF.<br /><br /> Of course, ALL the nodes must be in contact with each other not allowing
signficantdrift on their clocks while operating. (NTP is a starter)<br /><br /> I just do not know of any other way to
addself healing without the above information, regardless of whether you hold up transactions for synchronous <br /> or
letthem pass thru asynch. Regardless if you are getting your replication data from the WAL stream or thru the client
libraries.<br/><br /> Also, your replication model does not really discuss busted link replication operations, where is
theintelligence for that in the operation diagram?<br /><br /> Everytime you package up replication into the core,
someonehas to tear into that pile to add some extra functionality, so definitely think<br /> about providing sensible
hooksfor that extra bit of customization to override the base function.<br /><br /> Cheers,<br /><br /> marco<br /><br
/>On 9/22/2012 11:00 AM, Andres Freund wrote:<br /><blockquote cite="mid:201209221900.53190.andres@2ndquadrant.com"
type="cite"><prewrap="">This time I really attached both...
</pre><br /><fieldset class="mimeAttachmentHeader"></fieldset><br /><pre wrap="">
</pre></blockquote><br />
