<html>
<head>
<meta content="text/html; charset=windows-1252"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<br>
<div class="moz-cite-prefix">On 06/12/16 12:15, Andrey Grebennikov
wrote:<br>
</div>
<blockquote
cite="mid:CAP2HsxNZ8WwVD94mAf7ZfuX8KqqJRGPXOq+jgrfumCE6CF8+DA@mail.gmail.com"
type="cite">
<div dir="ltr">
<div class="gmail_extra">
<div class="gmail_quote">
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
<div bgcolor="#FFFFFF" text="#000000">
<p>Just to clarify, the reason you want custom project
IDs is so that when you create a project in one
region, you can create it again with the same ID in
another? Isn't that just manual replication though?<br>
<br>
What happens when there are projects in only one
region? Or if that isn't meant to happen, how would
you make sure that doesn't happen?<br>
</p>
</div>
</blockquote>
<div>Not quite sure I understand the point...</div>
<div>Yes, this is exactly the reason why I need this
feature. Project creation/replication is supposed to be
done via external scripts. Currently I already have 2
slightly different use cases which definitely require this
feature.</div>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex">
<div bgcolor="#FFFFFF" text="#000000">
<p> </p>
<br>
</div>
</blockquote>
</div>
</div>
</div>
</blockquote>
My point was how is this any better than database replication? It
pretty much sounds like replication but at the API layer.<br>
<br>
Do you just create a project normally in one region (keystone API)
and then a cron job or something triggers the replication script, or
do you have a queue of changes to replicate? Isn't that much the
same as doing asynchronous database replication? Or are you creating
and duplicating in the scope of the same script? If the latter, then
that sounds like synchronous DB replication.<br>
<br>
How are you handling conflicts? Are you?<br>
<br>
If networks are down, the same problem is present. If ProjectOne is
created in RegionOne, and ProjectTwo in RegionTwo, until the
inter-region network is restored, the two keystones will be
different until your script runs in both regions and actually gets
through to the opposite one.<br>
<br>
What is the benefit to doing things this way rather than just
trusting your replication to the DB layer?<br>
<br>
I'm genuinely curious about your approach as we've been looking at
moving to a multi-master like set up for our keystone so I'd love to
know what problems you found with DB replication. API level
replication sounds like a terrifying idea, so I'm curious what
exactly your reasons behind wanting to do it are.<br>
</body>
</html>