12466
Offtopic group for casual talking about anything. @rules_for_python still apply (except for the ontopic rule)
Before joining this group I failed to solve CAPTCHA and was blocked by Dorogonkiraa in an on-topic group.
Читать полностью…
Hope things are going well, how much growth has your semiconductor business achieved so far?🤔
Читать полностью…
I think Taaz is a boss I mean owner, so maybe tagging her without a username won’t work or she may not notice.
Читать полностью…
Hello sir. I know several people in india. What's your skill stack? Tensorflow? Pandas? etc. What do you know right now to land a job. Market is very competitive right now sir.
Читать полностью…
I listened til 20 min then got into meetings. That’s good to hear
Читать полностью…
However, if this is not about hosting but secure communication, let them provide the VPS and setup a Matrix network.
Читать полностью…
Blobs, media, backups: Garage S3
Databases needing HA: App-native replication or accept node-loss risk
Databases tolerating downtime: Local storage + S3 backup + restore on failure
Stateless workloads: Anything, freely schedulable
Still talking about stateful applications. Stateless is trivial.
Читать полностью…
I did a three nodes k3s cluster with gluster as backing storage for PVCs, across three different data centers that were "near" each other. I didn't run much on it though.
Читать полностью…
Vault controls how secrets are distributed and who is authorized to request them — but once a secret lands on a node, the node operator can see it.
Читать полностью…
Longhorn is slow and synchronous. Garage is a good fit for distribution data over nodes that are not near to each other.
Читать полностью…
Yes, I just tried. Single verification process apply to all groups.
Читать полностью…
It is not a semiconductor business, it is an Embedded System business... it is going fine, but I'm sure there will be a wrench in the engine by our great politics.
Читать полностью…
Tazz, which model should be selected as sonnet substitute?
Kimi 2.6 maps to Opus
Qwen 3.6 35b -> Sonnet?
Haiku -> devstral small 2 24b?
Sub agent model -> some fast model?
Hi everyone, how are you effectively using AI models in cursor to maintain the request quota
Читать полностью…
How can I land avg IT job as ai engineer please enlighten me
Читать полностью…
I found the start wasn't so good, but I like it now. Into it an hour.
Читать полностью…
Think about this: If they paid you for one VPS and you would rent it, they would not have root access on that node, but you could grant them secure access to all nodes inside Kubernetes using RBAC and limit what they can do per namespace: CPU requests and limits, memory requests and limits, number of pods, services, PVCs and storage capacity. With CNI like Calico (that would replace the in-built Flannel from k3s) you could also make sure that cross-namespace communication is restricted and bandwitdh is (partially) limited as well. Cilium might be a better fit for limiting bandwidth even, but it's a bit more complex.
From a security point of view, that'd optimal.
The workloads that are supposed to run on the cluster need to be considered now to be able to make a good decision about the storage requirements.
Читать полностью…
For the scale you are planning for, S3 is a good fit but you need to say goodbye to distributed storage that runs on those nodes and rely on the applications running on those nodes to replicate and create and upload backups. It's not fully self-healing when a node goes down.
Читать полностью…
Well, that's a task to think about. I will design it when there are volunteers
Читать полностью…
Longhorn integrates with k8s. Its could be a nice solution
Читать полностью…