> WE'RE GOOD JOURNALISTS [subliminal messenging from MIT]
> ADS
> GIB US MONET?
>
???ah good i have rarely until now considered "reader mode" to be necessary.
maybe i should finish porting configs over
i don't think i will bring it back though?
probably, i will just make a fish script that copies dotfiles to a staging folder and another fish script that copies dotfiles from the staging folder to their proper homes.
then i dunno. pyfabric (
) or sshkit to trigger the deploy (
)
🇳🇱
@icedquinn May not be what you are looking for but is this really just for backup so you can restore to a point? If so maybe consider incremental backup. I use BTRFS for filling that need. My main drives are all ultra high speed raided NVMe drives and then i have one slow but very large (12TB) spinning metal hard drive also running BTRFS that I backup to.. Makes restoring to a point or just restoring individual files trivial and backups take up minimal resources.
Of all the solutions I've tried, including ones specific to my configuration directories (like git) this has turned out to be the least painful.
there are also other older mostly decom'd computers here that i just never got rid of. it's nice to boot them on rare occasion to make sure they still work in case something really bad happens.
@icedquinn Ahh so sharing configs to multiple computers.. ok yea thats a very different use case than I envisions. BTRFS isnt likely to be so helpful there.. yea I'd say ansible and/or chef are your two top options then
@freemo most important is just fucking doing any of it which i’ve been putting off.
@icedquinn Well glad you are feeling productive then!
@icedquinn havent messed with salt yet.
Kerberos is ok, but i find the syntax of its files, particularly since it is proprietary, to be kunky, and integration with more standard formats are limited at best. I use Swarm myself for a similar purpose.
In fact all the QOTO servers are running on swarm.
not familiar with swarm.
there is also apache directory server, and some people use freeipa.
i just used the base MIT daemon because *it was there*.
ansible is primarily running a linear series of commands on a bunch of servers that it cycles through.
salt tries to enforce "states" which in turn gets farmed out to local python scripts which interpret what that means for that particular machine. it is also built off an RPC mechanism that communicates with a master.
either one can trespass as the other (ansible has a pseudo-master, salt has masterless) and you *can* do a salt orchestration file that just runs a bunch of shell scripts, but.. yeah.
ansible is cheap and easy to train, salt is what scales. they're only semi-competitors. especially since you can use salt entirely as an rpc and monitoring system and not provision packages with it if you choose. for instance, you can target a salt-call to only EU servers and then tell them all to drop their vault instances immediately.
i had briefly set up a little tiny kerberos domain and it was kind of interesting to have ansible install the domain configs to various computers.
but it's an entire toolchain i only need for the purpose of copying a couple files, so i dunno. might compare the load to have ansible installed and remember how to write playbooks to the load of sshkit (which is basically just an ssh runner for rakefiles.)