- Joined
- Jan 1, 2016
- Messages
- 9,703
OK, so I feel the need to speak up a bit on a topic that I've had on my mind since the summer of 2000.
I had a bit of time on my hands and was looking to get involved in testing/helping during the first real development window for Scale (angelfish).
At that time, I was absolutely bursting with enthusiasm about what SCALE was promising... it was exactly what I wanted (and probably more) in order to meet my simple dream (I think I even posted something like it somewhere in a thread with iX product managers, but can't find it easily now):
It sounded simple enough, although was a big gap I had found in the way that Rancher 2.0 did things with Kubernetes... but I saw an opening there with "host local paths" (or something like that), meaning each node would look for config in the same (local) path, so all that was needed was a bunch of nodes that kept a set of files available in the same path synced up for the apps to use and some kind of "app coordinator" (like Rancher) which would decide which node was best to run an app on.
What showed up was not that.
Gluster looked like a good idea and I persisted with that a bit, but never really got what I needed. Now it's gone altogether, never having really got airborne.
SCALE prevented running a more capable front-end like rancher to do the kubernetes clustering and replaced that with what was a very low-tech alternative which as far as I can tell still isn't going in the right direction 4 years later and has really put a lot of folks off with the catalog wars and the unfortunate events that unfolded with TrueCharts and their initial releases of changes requiring app re-install (and then some more).
One glimmer of hope I saw was the iX publishing of syncthing, but I remain skeptical without any promise of a GUI to manage apps to consume it in a sensible way.
I'm now considering options for how I might be able to get closer to this utopian dream using the systemd-nspawn container system, but have hit walls with running Kubernetes in those already.
For something that once had its own use-case, I think SCALE is trying to be CORE too hard and forgetting the original goals of being SCALE-out.
I look forward to what the future and dragonfish may bring, but I suspect I'll be waiting a long time to have my dream setup since I see nothing in the notes so far to indicate it is moving in that direction.
I don't want this to sound "rant-y", so I'll add that I'm really happy to have a company like iX looking after the product in a way that allows access to a wide audience for a wide variety of use cases. (but I still want my dream setup @Kris Moore :P ).
I had a bit of time on my hands and was looking to get involved in testing/helping during the first real development window for Scale (angelfish).
At that time, I was absolutely bursting with enthusiasm about what SCALE was promising... it was exactly what I wanted (and probably more) in order to meet my simple dream (I think I even posted something like it somewhere in a thread with iX product managers, but can't find it easily now):
It sounded simple enough, although was a big gap I had found in the way that Rancher 2.0 did things with Kubernetes... but I saw an opening there with "host local paths" (or something like that), meaning each node would look for config in the same (local) path, so all that was needed was a bunch of nodes that kept a set of files available in the same path synced up for the apps to use and some kind of "app coordinator" (like Rancher) which would decide which node was best to run an app on.
What showed up was not that.
Gluster looked like a good idea and I persisted with that a bit, but never really got what I needed. Now it's gone altogether, never having really got airborne.
SCALE prevented running a more capable front-end like rancher to do the kubernetes clustering and replaced that with what was a very low-tech alternative which as far as I can tell still isn't going in the right direction 4 years later and has really put a lot of folks off with the catalog wars and the unfortunate events that unfolded with TrueCharts and their initial releases of changes requiring app re-install (and then some more).
One glimmer of hope I saw was the iX publishing of syncthing, but I remain skeptical without any promise of a GUI to manage apps to consume it in a sensible way.
I'm now considering options for how I might be able to get closer to this utopian dream using the systemd-nspawn container system, but have hit walls with running Kubernetes in those already.
For something that once had its own use-case, I think SCALE is trying to be CORE too hard and forgetting the original goals of being SCALE-out.
I look forward to what the future and dragonfish may bring, but I suspect I'll be waiting a long time to have my dream setup since I see nothing in the notes so far to indicate it is moving in that direction.
I don't want this to sound "rant-y", so I'll add that I'm really happy to have a company like iX looking after the product in a way that allows access to a wide audience for a wide variety of use cases. (but I still want my dream setup @Kris Moore :P ).
