What are some small improvements that you did to a backend or frontend repo?

What are some small improvements that you did to a backend or frontend repo? Trying to figure out if there are little things I can do to make development better or faster for my team.

https://redd.it/1mt5klq
@r_devops

Someone created a DevOps version of Cards Against Humanity called "Clusters Against Humanity"

Im assuming many of you are familiar with the "Cards Against Humanity" game. Well, someone created a game called "Clusters Against Humanity" for Devops/Elasticsearch/Opensearch folks - looks pretty well done and thought the folks here would get a laugh out of it https://clustersagainsthumanity.com/

https://redd.it/1mtea9u
@r_devops

OCR struggling with Greek historical newspaper text - need advice on preprocessing/recognition techniques

Hi everyone!

I'm working on digitizing Greek historical newspapers from the 1980s and looking for advice on improving OCR accuracy for challenging text.

# What I'm working with:

* Scanned Greek newspaper pages (see attached image)
* Mix of Greek text with occasional Latin characters
* Poor print quality, some fading, typical newspaper scanning artifacts
* Historical typography that doesn't match modern fonts

# Current approach:

* **Tesseract with** `ell+eng` **language models** using various PSM modes (3, 4, 6)
* **Preprocessing pipeline:**
* Grayscale conversion + upscaling (2x-3x using INTER\_CUBIC)
* Noise reduction (Gaussian blur vs bilateral filtering)
* Binarization (Otsu vs adaptive thresholding)
* Morphological operations for cleanup
* **Post-processing** with regex patterns for common Greek character corrections

# Looking for advice on:

1. **Better OCR engines** \- Has anyone had success with PaddleOCR, EasyOCR, or cloud APIs (Google Vision, AWS Textract) for Greek historical documents?
2. **Advanced preprocessing** \- Any specific techniques for newspaper scans? Different binarization methods, contrast enhancement, or specialized denoising approaches?
3. **Training custom models** \- Is it worth training on similar Greek newspaper text, or are there existing models optimized for historical Greek typography?
4. **Workflow optimization** \- Should I be doing text region segmentation first? Any benefits to processing columns/paragraphs separately?
5. **Language model considerations** \- Better to use Greek-only models vs mixed Greek+English for newspapers that occasionally have Latin text?

**Context:** Planning to scale this to thousands of pages, so looking for approaches that balance accuracy with processing efficiency.

Any insights from folks who've tackled similar historical document OCR challenges would be greatly appreciated!

**Tech stack:** Python, OpenCV, Tesseract, PIL (open to alternatives)

# you may check an image sample from here [https://imgur.com/a/tVgHWFq](https://imgur.com/a/tVgHWFq)

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

What does AI really look like in data engineering?

You guys might have noticed… there’s a lot of hype about “AI-ready data stacks.” But, it definitely isn't simple to achieve. Freshness, reliability, orchestration, the bar is just different when theres LLMs involved.

After a lot of brainstorming and chats with industry experts, we set up a 45 mins webinar with Hugo Lu (ex-data infra engineer, now founder @ Orchestra) where he’ll share his take on how AI is changing data ops, and what pipelines need to look like when LLMs are involved

Its totally free so if youre interested or just want to know the implications of AI in your stack, do join us 🙂

Its on  Aug 21, 1 PM ET

Register here!



https://redd.it/1mtfxl7
@r_devops

How to handle version numbers when promoting Docker Images from SNAPSHOT to Release

I have a CI/CD pipeline split into two stages:

Pre-Production Pipeline:

Builds & tests my Java Spring Boot application
Creates Docker image with -SNAPSHOT version (e.g., myapp:1.23.7-SNAPSHOT)
Deploys to Docker registry for testing

Release Pipeline:

Should promote the tested SNAPSHOT image to production with clean version (e.g., myapp:1.23.7)
Creates Git release tag
Deploys to production registry

The Problem: I want to follow the best practice of promoting already-tested artifacts rather than rebuilding. However, some files inside the Docker image contain version strings that will show -SNAPSHOT in production. For example:

`manifest.json` (visible in browser dev tools for frontend apps)
package.json
JAR manifest files
Any compiled assets with embedded version info

Current Environment:

Jenkins with Docker Pipeline Plugin
Nexus Docker Registry
Java Spring Boot backend
Angular frontend

The Core Question: How do you handle version string cleanup when promoting Docker images from SNAPSHOT to release without rebuilding the entire application?

https://redd.it/1mtip41
@r_devops

What's the next certification to goal?

I obtain the KCNA certification two weeks ago and i think CKA is the next certification to goal. It's natural course the CKA certification and i have pratical experience for one year and half with kubernets. Althought no provide the cluster, just delivery the applications by argocd and gilab. It's the better choice CKA exam? Or it's better investing in some cloud certification? I mean, not about the knowledge but the profile improvement, it's better choice Kuberstraount follow or some cloud provider like AWS to goal?
I have solid background like as sysadmin and obserbability, but still not considered myself a specialist.

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

What’s your go-to tool for monitoring Kubernetes clusters?

I’m managing a small Kubernetes cluster and struggling to get good visibility into resource usage and pod health. I’ve been using Prometheus with Grafana, but the setup feels clunky for my needs. What tools do you use for monitoring your K8s clusters, and what makes them stand out?

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

Winscp using ADSSHlogin extension

is there anyone who tried to use Winscp to connect to a linux VM using the ADSSHlogin extension?

I can use correctly the ssh command with the following commands:

set /p ipAddress="Enter IP of VM you're connecting to: "

call az login --use-device-code

call az ssh config --ip %ipAddress% --file %userprofile%\\.ssh\\sshconfig

call az ssh vm --ip %ipAddress%


but then I try to use the private key generated to use winscp wihtout any success

https://redd.it/1mtp2ke
@r_devops

My journey in building a GNU/Linux aarch64 (ARM) system

In these times of rapid technological change, I feel the need to go back to the roots... That’s why I started this project — partly as a personal exercise to refresh my fundamentals, but also to create a solid foundation for the upcoming DevOps & SRE Academies I’ll be organizing in the next few months at my work place. Here are some notes from my journey of building a GNU/Linux aarch64 (ARM) system: https://devopstribe.it/2025/08/18/my-journey-in-building-a-gnu-linux-aarch64-arm-system/

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

how much do you trust AI with deployment?

what the title says.

i feel that deploying nowadays feels either too magical (but locked into a certain rigid framework) or too manually to the point where you're better off just going straight to a cloud provider.

As these tools start developing their own alternatives, I'm curious on how other engineers feel about AI doing deployment for them. Would you ever trust an AI tool to actually deploy an app or infrastructure end to end? If not, why? When would AI be good enough to trust it like that?


cheers!

https://redd.it/1mtuy5f
@r_devops

Pivoting back to cloud-native development and DevOps

Hey everyone,

I’m looking for some insight and maybe a bit of guidance. I have \~5 years of experience in software development, most of it in cloud-native development (mostly with Azure). I have a couple of AWS certs (Solutions Architect and Developer Associate) and used to work in a role where I was pretty deep into the cloud stack—containerized workloads, IaC, CI/CD pipelines, microservices, etc.

Earlier this year, I made a career move into embedded systems for better pay and what I thought would be an interesting change of pace but I’ve found I really don’t enjoy the work. It’s very far removed from what I liked about building and deploying in the cloud—things feel much slower, less flexible, and I miss working with modern DevOps tools and practices.

Long-term, I’m aiming to become a solutions architect, ideally on the cloud side. I’m trying to figure out the best way to pivot back into that world.

A few questions for anyone who’s been in or adjacent to this path:

1. What’s the current and near-future outlook for cloud-native development and DevOps roles?
2. Is there still strong demand for people with cloud experience and certs?
3. Any advice on how to make myself more marketable again for cloud or DevOps roles?
4. Would aiming directly for a junior/pre-solution architect position be realistic, or should I try to get back in as a cloud engineer/developer first?

Appreciate any thoughts or experiences you can share.

https://redd.it/1mtwc8j
@r_devops

I got offered a dev ops role over help desk. Should I take it even though I can fail?

I’m more of a jr sysadmin. Went to school for computer science. Have programmed little since graduating few years back. The majority of my work revolves around help desk, and I’m looking to break away from that once and for all.

For two months I asked the security department if I can work with them on some tasks because I want more things to put on my resume and learn.

Well they know I want to get out of help desk related tasks so they offered me a role where my main responsibility would be using puppet to do configuration management on the servers. I was apart of the puppet training. I would also be doing other security related tasks.

I don’t do much programming, but should I accept this role? Can I learn most of it on the job? To tell you the truth I don’t want to be on the operations side of things much anymore. I really regret not doing programming outside of school.

Basically should I take this opportunity? I don’t think my pay is adjusting but I much rather do this role, it’s just I’m starting from not much experience. But I feel like I’ll be a lot more motivated to learn and better myself outside of work because I have a clear path.

https://redd.it/1mu2osm
@r_devops

How to make a dead project alive?

I'm a DevOps engineer working for a US based telecommunication company. We've been using CISCO as our vpn provider for years now. Apparently, it looks like CISCO is having licensing problems and we cannot go ahead with it for the long run. Before I joined, the previous engineering manager suggested that we should use Nordlayer as a replacement for everything we do with CISCO. He made a plan, convinced everyone that it'll work and suddenly he left the company. Now in the DevOps team it's only me and newly joined manager. Other people at company mentioned that previous manager had all the poc setup in our AWS, but as I checked it's not.

Basically what we want to do is make a connection between our DC to AWS via Nordlayer!

There's no documentation on how this thing worked previously, but management tells us that it needs to be up. We contacted Nordlayer support and they also do not have any documentation since it was a poc setup. So we're kinda stuck and the heat's on me because I'm responsible for AWS and previous setup was present in AWS. So I'm really not sure what needs to be done!
Thought of posting this here because I'm sure everyone must've gone through this situation once in there career to finish what's been dead for years. Help me out.

https://redd.it/1mu6f7p
@r_devops

Docker Setup for App with Frontend + 2 Backends + Certbot

Hi there,

I want to have a "simple-to-maintain" setup for having the following setup:

* Certbot
* Nginx (as reverse proxy sending traffic to backend & frontend)
* Two backends
* Frontend (Angular app)

For now I would do the following:

Certbot on the host without docker. To me it seems that having it inside docker-compose is too much of a hassle. The setup will be standalone so there will not be any other services requiring is to be in an isolated environment.

Nginx: To me it seems that it does not offer much benefits to put it into docker?

Backends: dockerized apps Inside a docker-compose

Frontend: How would you serve it? Would you rather put it on the host (since nginx is on the host already) or dockerize it (if so how?)?

What's your take on the setup?

https://redd.it/1mu9vir
@r_devops

Best Course+Certification for Devops

I am a currently pursuing BTech CSE and learning cybersecurity.

1. Want to know if DevSecOps is a viable field to get into in the future.
2. I will definitely need to learn Devops whether i choose to get into devsecops or not. Can someone please suggest some certification that i can aim for, preferably with a course accompanying it.
3. Which Platform do you think i should learn as a cloud platform(AWS/Azure) and Container Tools(Docker/Kubernetes), as i believe they both go hand in hand(correct me if i am wrong).

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

Is my plan valid

So I am actively learning DevOps related skills such as tools methodologies, hands on practice labs all by myself. I don't really have any job experience. Though I am currently doing a devops internship.
Which has almost 1.5 months remaining to complete. After that internship I work remotely as a freelance data scientist where my friend gets all the projects for putting food in the table so I am working almost 16 hrs daily.

This routeen is very taxing on my health so.

My Plan
After finishing my internship I plan to keep doing the freelance work, and then keep learning DevOps related stuff by myself as I have a clear direction now. To cater the experience issue I plan to get certs.
The cloud practitioner
Solutions architect
And maybe the sys admin professional.
And I know I will not just preparing for the cert but actual practice first and cert exams at the end.

So what do you guys say.

Getting a relevant job at the place I am an intern at might be a possibility but it does not seem likely. Even if I do get a job there, they use old styles stuff on prem servers, windows servers vms. Somehow it is related too.

In short: this is a bad company pays less , minimal learning.

Thanks

https://redd.it/1muc6s0
@r_devops

Inferencing GPT-OSS-20B with vLLM, Observability for AI Workloads

https://www.parseable.com/blog/vllm-inference-metrics

https://redd.it/1mufq5r
@r_devops

Which is the best Book of Networking for DevOps?

I am on the way for DevOps and now I want to learn Networking but have no idea which book should I read that should be sufficient for DevOps. As networking in itself is a very large topic so I was hoping for only What is necessary for DevOps.

https://redd.it/1muie59
@r_devops

Saving audit logs

Hi,
How can I best save audit logs for a company?
I tried using Grafana with BigQuery and GCS archive. The storage cost in GCS is cheap, but the retrieval fees from GCS are very high, and also BigQuery query costs add up.

Any advice on better approaches?

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

Anyone else hit a wall with CI/CD pipeline bottlenecks?

Last week, our team’s CI/CD pipeline started choking during a big release. We’re using Jenkins with a bunch of custom scripts, and it took hours to debug why our tests were hanging. Turned out, a misconfigured Docker image was clogging the build queue. We fixed it by pruning old images, but it’s clear our setup needs an overhaul. Have you dealt with pipeline bottlenecks like this? What changes or tools helped you streamline your CI/CD process?

https://redd.it/1mur0f9
@r_devops