deploy MiniCPM-V / Florence-2 to ec2


How can I deploy MiniCPM-V 2.6 ( https://github.com/OpenBMB/MiniCPM-V ) and Florence-2 ( https://huggingface.co/microsoft/Florence-2-large ) on ec2?

Which instance should I pick?

https://redd.it/1ep1xxp
@r_devops

How To Setup Self Managed Redis Sentinel on Prod Environment

So i was doing POC for self managed redis-sentinal. It seems to be working fine, 3 sentinal nodes, 3 redis nodes. HA and Auto-Failover was working smoothly on docker but when i treid the same approach with multiple ec2 and automated it through Ansible, in sentinal master node, i was able to see no of slaves 2 but on the slave nodes, the master_link was showing down. I don't know if it's due to using multiple nodes or something else but all the 3 ec2 were able to connect to each other.

The second Problem will is that, in all the videos i saw online, when the master redis goes down, another redis becomes master thanks to Sentinel but those guys in the YT videos were manually changing the master node ip address in the app. In production environment we can't do this, right? Then how can we achieve this like if somehow i can create a middleware link which help connects the application and redis master node. If any one of you guys has done something similar, please let me know. Thanks!

https://redd.it/1ep8q8m
@r_devops

Devops vs SWE internship

Hello,

I am about to start my final year as a computer science student and this summer I have been working on getting as much experience as possible.

I have landed 2 internships, one as a C++ SWE Intern and the other one as a DevOps Intern (mostly oriented on Python Bash, Linux, Kubernetes and other tools).

Both offers are identical, same company, I just have a really hard time to make a pick since I kinda love both.

I've been loving C++ ever since high school and I've always wanted to get to work with it.

In the same time, I have discovered other things I also find interesting. I finished 2 small practice programs where I had contact with DevOps technologies at a more experimental level (Github Pipeline, Docker, Jenkins, Kubernetes etc.)
I am also currently pursuing the first 2 Cisco CCNA certificates which I assume they are more related to a DevOps career.

In short terms, I love C++, but I am wondering if it would be more helpful to focus on a role that is focused on something more than pure development. I've been really discouraged by the job market in Software Development/Engineering and I feel like it's good to know more than just coding.





https://redd.it/1ep8c4u
@r_devops

finding OS level virtualization solutions

I want a simple sandoxed environment where i can open shell, install stuff, test webapp, as i would in any other normal OS, and sync that local sandbox to remote server.

I tried Docker, Chroot, lxc, Freebsd jails and had problems with them all

Docker - I don't like because it has forces you to use some magic yml configuration language - edit Dockerfile, docker-compose.yml. I just want to open up a sandboxed shell and run some shell commands.

Chroot - I got tired to fiddle around with consistent permission and mount errors

lxc - i'm getting some bs "userns_exec_mapped_root: 4887 No uid mapping for container root" error which i wasn't able to google any solutions

Freebsd jails - well, maybe its a best option so far, problem is that i don't want to install it on hardware because of lack of drivers and also not many host providers give this bsd as server OS option, at least AWS has it, which i'm probably gonna use anyways. I'm having some networking errors and vm always complains that disk is full even tho i grew it +20G, so skill issue i guess.

Am i missing something? I can't seem to find a working solution that fits.

related articles:
https://en.wikipedia.org/wiki/OS-level_virtualization
https://kevinboone.me/containerfromscratch_chroot.html
https://wiki.alpinelinux.org/wiki/Alpine_Linux_in_a_chroot
https://linuxcontainers.org/lxc/getting-started/
https://docs.freebsd.org/en/books/handbook/jails/

https://redd.it/1epf8f2
@r_devops

🦎 Monitor v1.13 - Komodo Release 🦎 Open source build, deployment, and automation platform. I'm now hosting a demo, check it out.

🦎 Demo: https://demo.monitor.dev

🦎 Docs: https://docs.monitor.dev

🦎 Release: https://github.com/mbecker20/monitor/releases/tag/v1.13.0

Hey guys,

Monitor is a platform to make declarative infrastructure easily. It breaks down the components of the CI/CD pipeline into Resources, like "Server", "Build", "Stack", "Alerter". All of these components and configuration can be delared in TOML files, checked into git repos, and synced to Monitor on button push: https://raw.githubusercontent.com/mbecker20/monitor/main/screenshots/Light-Sync.png.

It can build docker images using the "Build" resource, deploy docker containers with "Deployment", deploy docker compose stacks with "Stack". You do pretty much anything with the "Repo" resource, which gives you control over cloning a repo and running a shell command in the repo as CWD.

More info on Resources and syncing here: https://docs.monitor.dev/docs/resources#resourcesync

It can spawn single-use AWS instances to run tasks like - Clone repo run command, or docker build and push image.

Everything is open source and the backend is all in Rust.

🦎 Hope you enjoy 🦎

https://redd.it/1epg5ky
@r_devops

DevOps to Pipeline Engineer: Need resources for GitHub Actions best practices and optimization


Hi all, i worked as a devops couple of years, but in my new position, i was tasked to help developers with their github actions pipelines. I am not very proficient on it because in all my past devops roles, i was mostly into aws, (cloud engineer) and this company is a bigger one, they got cloudops for the aws things, they got sre's for kubernete and monitoring, so this devops team are actually a "pipeline" team.

The thing is that developers are actually very capable of building their own pipelines, so every time i get on a call with them they ask "so, what can you do for us" and i actually don't know what to respond..

Could someone point me into a book or something that can help me compile a checklist of best practices that a pipeline can have? for example, i was reading a nice application security book that in the end it had checklists of things need to perform and i am sure there is something similar for pipelines that i just can't seem to be able to find.

Through personal research i found that there are things like scorecards, sboms, attestations, container signing etc, but i am sure there are more like "how to make a pipeline run faster" etc. So i would appreciate if someone could point me to the right direction, thanks!



https://redd.it/1epg2ly
@r_devops

SpecFlow UI Tests Failing on Azure Pipelines Despite Extended Wait Times and Workarounds – Need Advice!


I'm facing a frustrating issue with some of my UI SpecFlow test cases running in Azure Pipelines. They keep failing with a NoSuchElementException, even after trying several fixes.

I've already:

Increased the ExpectedCondition wait from 12 to 30 seconds.
Changed the implicit wait polling from 1 second to 3 seconds.
Added a hardcoded wait of 3 seconds before interacting with the element.
To address this, I suggested a few changes to our DevOps team, and they were implemented:

Added test run in parallel.
Enabled failedrerun as true.
Set UI test to true.
Unfortunately, none of these solutions have resolved the issue, and the tests still fail intermittently. I suspect it might be due to the slower performance of the pipeline server/MS agents, but I can't seem to find a solution that works consistently. Has anyone else encountered this issue or have any suggestions on how to tackle it?

Any advice would be greatly appreciated!

https://redd.it/1eph3hf
@r_devops

Azure vs. AWS

If you had to pick on of these cloud providers for a long run, which one would you pick and why?

https://redd.it/1epk8mz
@r_devops

Strategy/Tech for managing/building docker images on a larger scale?


Hi, we are slowly moving towards more widespan use of containers for our applications. Already seeing way faster turnaround on dev-cycle and all the fun DevOps stuff.

Anyways! We are a on-prem company that have some the usual quirks that follow. For example proxies here and there in the network and SSL inspection that needs its certs to be shimmied into our images..

So to make the process easier the devs we have "shimmed" a bunch of the basic official images. IE add the CA certs and the config that follow, add the internal artifact repository and the likes.

The images are currently built with yaml shenanigans and docker-compose, but I'm not really all that happy with the setup.. Even more so as the amount of "base" images we maintain grows..

So, anyone got any good tips for managing a growing volume of images?

https://redd.it/1epk2t7
@r_devops

Microservices into an on premise installer?

Hello, I would like to ask if any of you could help me with ideas how to achieve packaging microservices into a distributable offline installer.

The idea is that the customer installs this in their own server with a couple of clicks, but the end product consists of microservices so they have to be "installed" and managed somehow. I am not allowed to use docker but I would be open to ideas using that

https://redd.it/1epnd2y
@r_devops

I might need to "breakout" of DevOps

So, I am a DevOps student and I have been studying for 3 months since trying to "break in" and I'm starting to feel discouraged, or more like the field is too "technical" for my interests, so I need advice on what I could do to help (or if I'm a lost cause).

Context: my background is in medicine and I have been interested in joining the tech/computer science atmosphere since literally finishing my BS in medical professions. I didn't really want to become a doctor like I set out to be maybe 2 years into undergrad, but I still loved stem and wasn't really sure of my other interests, so I continued and still kept my major (also was young, burnt out, and not interested in spending more time in school). For the past 3 years, I really observed the tech field and thought I found myself really interested in joining (due to other peers making switch from medicine, pay, and I started super basic level coding and thought it was pretty fun). Maybe, the boosts to the field during COVID sparked my interests too. I enjoy the lifestyle and work life that comes with tech and prefer alone work rather than team work.

Now, I am 3 months into a studying devops, I've learned (but not mastered) Linux, AWS, Ansible, and GitHub, and barely any Docker, but I just feel so lost. I'm not sure if it's my study methods but I feel like I'm not going to do well in the field or I'll just be super lost (and that scares me!!!!). I have no idea how to answer interview questions, and I feel less passionate about learning and studying more since I feel like I'm not really grasping everything "technically". I feel like this field is extremely hard to break into if your technical background was non existent before studying like mine. What do you guys think? Should I take my interests to another tech field?

https://redd.it/1eple0x
@r_devops

Can’t resolve a sonatype vulnerability

Every time I run my pipeline in azure devops, sonatype picks up a vulnerability for a dll file that needs to be updated versions. I update the file to the needed version and then run the pipeline but the file keeps going back to the version that is vulnerable. This file is in the Release folder

How do I fix the file version permanently? This is a .net project

https://redd.it/1epq787
@r_devops

Need Guidance (Student)

Hey, There I am an engineering student, I have a pretty good interest in DevOps. I have started some lechers about AWS, if there any small amount of guidance you can provide me that would be helpful. For getting started

https://redd.it/1eprn6m
@r_devops

What do you think of platform engineering?

Is it the next generation devops?

https://redd.it/1eps6p9
@r_devops

Is there a book/course/podcast/tool-suite that helped you go from fundamental understanding to mastery?

15 year software developer entering devops space.

https://redd.it/1epsrj2
@r_devops

How do you handle disaster recovery?

As a best practice, let's say.

And on that note, what does HA mean to you?

https://redd.it/1epvu41
@r_devops

Distributed Tracing Weekend Project with Grafana Tempo

🌟 This weekend, I got curious about exploring Distributed Tracing in a Microservice architecture using containers in my Homelab. I decided to dive in and simulate an order processing system with multiple services: order-service, inventory-service, payment-service, warehouse-service, and fraud-service.

🔍 One of the challenges with microservices is debugging when a single request traverses multiple services. It can be tricky to trace and understand what’s happening across the system. To tackle this, I implemented distributed tracing using OpenTelemetry, Grafana Tempo, and Grafana Loki. The setup is designed to provide a seamless way to view traces directly from logs, making it easier to debug and monitor the entire process.

🚀 The project includes Docker Compose with auto-configuration, so you can easily spin it up and explore the architecture yourself. If you're interested, feel free to check out the repo, and don't forget to give it a ⭐️ if you find it useful!

Github Repo:

https://github.com/ruanbekker/grafana-tempo-loki-tracing

https://redd.it/1epww6b
@r_devops

Risks of running 2nd Express server with health check port?

I have a simple app running on NodeJS/Express ubuntu backend on AWS ec2, with free monitoring using UptimeRobot (UTR). I decided I didn't want to leave health check API exposed publicly, so I stood up a second express instance in my server.js, a second port (4431), and configured port 4431 to host only my healthcheck route. I then locked down p4431 access via Security Group to only IP ranges owned by UTR (https://uptimerobot.com/help/locations). It all works as intended, UTR can monitor successfully while the port and health check remain publicly closed. Just curious: Are there any risks or critical tradeoffs with this approach? Something like "a second express server drastically increases resource consumption", etc?

https://redd.it/1epy6cw
@r_devops

DevOps Testing Tools For 2024 Compared

The article discusses various testing tools that are commonly used in DevOps workflows. It provides an overview of the following popular tools for different types of testing (unit, integration, performance, security, monitoring) to help choose the right testing tools for their specific needs and integrate them: [9 Best DevOps Testing Tools For 2024](https://www.codium.ai/blog/best-devops-testing-tools/)

* QA Wolf
* k6
* Opkey
* Parasoft
* Typemock
* EMMA
* SimpleTest
* Tricentis Tosca
* AppVerify

https://redd.it/1eq4pyv
@r_devops

Pragmatic scaling of small self hosted CI runner fleet

Hi,

I'm managing the self-hosted CI infrastructure for a small software dev team. Mostly, I'm ensuring we have enough runners for the team needs. We don't have apps, the runners are just building code, running tests, etc (in docker, so the runners just need to have docker).

We have a couple small servers, and what I had so far was a basic Linux distro plus a homemade script to have the runners up and running.

Now I'm facing two issues : we're leaving gitlab for github, where runner can only execute one job, so while having the capability to parallelize a dozen jobs was just one simple param in a config file for gitlab, now I actually need to instanciate a dozen runners. Plus, the team grows, the CI does more and more, so I need to be able to add runners from time to time.

Now I'm looking for the most pragmatical way to scale this runner fleet up, given that I've never played with k8s, proxmox, ansible and the likes. It should be easy to maintain and scale, and not too hard to setup.

I'm thinking about getting proxmox on each node with 4 VMs, each having a runner (setup through a script), but managing all this manually already feels hard.

What are the best options, simple and not overkill, yet efficient, given that I don't know yet about k8s, we don't have a cluster or anything like that?

https://redd.it/1eq7sqy
@r_devops

HAProxy and hotsport on eks

Hello everyone,

I have a situation at work where I need to change how we expose our sandboxed environments to clients. Our current infrastructure runs on eks and we provision pods to clients on-demand using nodeport as the service type with one node in the cluster exposed publicly and acting as the entry point for the client connection. We are running this setup because all of the client connections are tcp based and the guy who designed the original infra obviously hasn't put much thought into the user-base growing and the nodeport range limitation posed by eks that we'd eventually run into (Only 2767 ports could be used simultaneously).

Now, I am thinking about using HAproxy controller and hostport to map the client connection directly to the pod, but I have no idea or how that would work, it's just an idea that I have initially. I would love to hear some solution suggestions and/or pointers on how I would start implementing the idea that I have. All the application pods are tcp based and I need to make an exclusive pod for each client those are the only two constraints.

https://redd.it/1eqayb4
@r_devops