I’ve been developing a tool to easily manage large numbers of vaults (eg on aws, azure, gce, digitalocean, local machines and vms etc). It’s basically a nice wrapper around a bunch of sysadmin provisioning and configuration tools, specifically tailored to vault info and operations.
Since I’ve already done a lot of work with provisioning and managing private safe networks I have a very good idea of what is needed for this tool and how to make it simple to use.
However I know market feedback is very important so if anyone has any particular requests for features please reply to this topic so I can incorporate them early and make the tool maximally useful for the people who will end up using it.
The tool is still a long way from being even a minimum viable product, but that’s the best time to gather this sort of market research.
I’ve got one too that I wrote for aws by wrapping their cli. Haven’t maintained it for about a year now so it could be broken. I was thinking it might be good food for thought when it came to building a “SafeCluster” computation app, but your idea sounds better for a first phase effort. PM me if you would like to collaborate.
Please, can you add vault options to target a specific section at a specific age? (just kidding, I know that Maidsafe will make sure that it won’t be possible)
Hey @mav, this sounds interesting, I presume specially for those who would be running many vaults, e.g. data centers. Just trying to understand a bit more and being curious, wouldn’t be enough or good enough to just use things like Kubernetes/Docker Swarm with vault docker containers? (I’m not an expert in using these tools I just happen to known them from previous jobs and used docker just a bit). And isn’t it also related to this other tool mentioned in this thread? I’m just trying to understand the differences with what you are working on.
Probably. I’m only slightly familiar with containers. The things that prompt me to go this route are
known hardware configuration correlates to known performance characteristics, which is important for profiling and also important for maintaining vault rank in a live network.
vms have simple network interfaces and thus simple configuration management; I’m not that familiar with containers so maybe it’s easier than I realise.
not all clouds support containers so this seemed able to target more platforms with less dependencies.
Sounds similar but I’m not familiar with ServiceFabric.
For me, the ‘abstraction’ offered by containers or vms or whatever is the least valuable part of the management of a set of vaults. The real value is in
managing the state of the network (eg total vaults, healthy vaults etc)
managing the ongoing reliability of the vault environment (cpu ram hdd etc usage)
predicting and monitoring the costs of the vaults
obtaining, interpreting and responding to vault logs
decommissioning vaults
containers don’t prevent any of these or even make them harder or easier. So I guess in theory there’s no reason why it couldn’t target containers.
When I first read what mav is doing, it seemed to be a lower level implementation, than what I’ve mostly seen implemented in ServiceFabric. But I was also immediately thinking about how it could be solved within this orchestration framework. After all, it actually manages high availability and optimum usage of node capacity etc., and a few of those points you mention @mav, it seems to me could be sort of propped up with ServiceFabric, or even made a lot easier, but if you know how to roll your own solutions I would say it’s nice and good to get it sleek.