@deepumohanp@devops_rob Nice fix. Sending PR screenshots to your own store instead of GitHub artefacts gets round the 500MB cap, and running it on Nomad keeps it next to everything else. If the 3-day cleanup isn't already one, a periodic Nomad batch job is a neat way to run it.
In Nomad a region is the cluster, a datacenter is a named scheduling domain inside it, and a node pool groups clients by intent. Datacenter and pool are separate filters that combine at placement. Omit node_pool and you get "default", not "any". https://t.co/XuU7V1Jvez #Nomad
@deepumohanp@grok We build Nomatron, a management layer for Nomad: approval rules on deploys and control over who can change what. Plain Nomad on one box is still a fine way to run things.
@HugoValters Nice. Worth pairing with the event stream API (/v1/event/stream) so you catch allocation and node-drain events as they happen, rather than reconstructing them after the fact.
@ohheyyang That's the big win solo. One habit that pays off early: keep jobspecs in git and run nomad job plan before every nomad job run, so the diff shows you what's about to change before it does.
@swapskp@alexxubyte Agreed. One twist that makes Nomad interesting in that comparison: it isn't container-only, so the same scheduler runs Docker tasks, plain binaries and Java side by side.
@shravanthreddyr Nice fix. A restart skipping shutdown_delay is exactly the kind of thing that bites mid-deploy and is hard to spot. Thanks for sending it upstream.
@ohheyyang Nice setup. One thing worth adding: a distinct_hosts constraint on anything running two copies, so both can't land on the same labelled box. A spread block is the softer option.
@lnxsec Agree. The same split applies one level up: being able to submit a deploy shouldn't mean being able to approve it. It's why service accounts in Nomatron can't approve deploys: where a rule needs approval, a CI token on its own can't approve its change.
@fullstacklover Nomad (no surprise from us). Real reason: it schedules containers, plain binaries and Java side by side, so older services can move over without being containerised first.