[Feature Request] Distributed JSS replication

Huge ditto on that one!! We have 24 sites scattered around the province, some with slow connections... there are a few big packages (like Adobe stuff) that we only need at one or two sites, but anything we put into the system gets replicated 24 times - ouch.

Jeff

(My apologies for mangling the subject line on my first post...Don)

Well having multiple instances is already possible. I haven't set it up, but I was thinking about having this for my JSS instances

FQDN.casper.com (as main JSS) FQDN.casper.com/mobile (ipad and MDM stuff)

We have separate people working on the iPad stuff, which is not really in my hands. I would like to set up an instance for them that only allows control of the iPads. Right now I just have them logging into the main JSS with some restricted access except for MDM access.

I would like to see the client cache all the JSS servers (like /etc/hosts or something similar) and then try each one if one fails or times out. Also, having ~1,000 machines per a site hit a single server instead of 8,000 hitting one server I think would load balance the check ins a lot.

just my 2 cents

-Tom

I've also a feature request with JAMF to allow for staged failovers.

I.e if a distribution point isn't up, it'll failover to the next & so on. So you can pick the failover that's applicable to your WAN.

Also, this failover system will work if a package does not exist on a distribution point (so if a policy tries to install a package from a point that hasn't replicated yet it'll failover).

Regards,

Ben.

yes on top of that I would like rules for cascading/replication. I would like to flag what packages/images can be replicated to what servers. So if I do have a different version among servers for an actual reason it won't get over written. I also don't need packages for primary schools going out the secondary schools distribution points. It makes the sync time faster and saves HD space.

-Tom

I think it's a great idea. It would also offer a layer of redundancy. If your parent went down, traffic could be routed to the child(ren), with the option to promote a child to parent and take over that role.

-Charlie

-------------------------------------
Charlie Smith
Desktop Engineer
Information Services Department (ISD)
MIT Lincoln Laboratory
244 Wood St. Lexington, MA 02420
Phone: 781.981.0854
E-mail: charlie.smith at ll.mit.edu
-------------------------------------

Similar to this idea would be what we've discussed before, which is
On 4/28/11 9:17 AM, "Thomas Larkin" <tlarki at kckps.org> wrote:
redundant JSS servers. Or at least the elimination of the single JSS
server as a single point of failure.

Whatever implementation is developed (master/slave or redundant), Casper
will never be an enterprise class tool until it at least offers redundancy
for the JSS.

--

William Smith
Technical Analyst
Merrill Communications LLC
(651) 632-1492

i imagine we're basically talking about tomcat clustering and mysql master-slave replication, or at least stuff along those lines.

it may be a good idea to use nginx or something else in front of all that for load balancing.

i'd like to see the mysql side optimized and the schema documented. i showed the db to a mysql db admin, and she almost ran away screaming.

I really would like database replication. I would like my clients to check into the nearest JSS instance based on VLAN location instead of all going over the WAN. The fail over would be the icing on the cake for me since I actually have an extra server lying around for a back up JSS. I would just have to have my network admin point the DNS to the new MAC address and I'd be golden.

Since I work for a school system we have tons and tons of online testing. It always kills our bandwidth. Maybe it will get better with Google fiber, I don't know, but some of our networking infrastructure is built off of older equipment as well. So that is always a limitation. We work on budgets, so we cannot always upgrade everything.

Thanks,

Tom

I totally agree !

Just adding something, it would be great to be able to select which packages get replicated to which replicas.

For example, all our designers & video guys are working in the headquarters. That mean I get the CS5 Design, Production and Web Premium replicated to all the child sites. It's a big waste of resources.

Francois

I would say that having two different packages with the same name in the same database (even on two different DPs) is asking for trouble. I don't think this aspect should be included, as it would significantly complicate replication, with little reason (just use two distinct packages with two distinct names).

On Apr 28, 2011, at 3:58 PM, Thomas Larkin wrote:

yes on top of that I would like rules for cascading/replication. I would like to flag what packages/images can be replicated to what servers. So if I do have a different version among servers for an actual reason it won't get over written. I also don't need packages for primary schools going out the secondary schools distribution points. It makes the sync time faster and saves HD space.

-Tom

I'd like to see time and day of week controls for the replication. It would be really cool if packages could be downloaded bittorrent style from other client machines. We don't have the luxury of super fast network connections between our locations yet, and our sites are often widely dispersed geographically.

Neal Costello
ICT officer
DPEM Tas.