Only one person can deploy
Releases go out over SSH by whoever remembers the order of the commands. That person is a single point of failure for the whole product, and bringing a second engineer up to speed takes a week of shadowing.
S/07 — devops and infrastructure
DevOps and cloud infrastructure, done end to end
DevOps and cloud work means taking an application that merely runs somewhere and making it run predictably: packaged in Docker, deployed from Git by a pipeline instead of by hand, watched by monitoring that alerts someone before customers do, and backed by copies that have actually been restored. We pick projects up at any stage — a hand-managed VPS, a cloud account nobody audits, or nothing yet.
Discuss your taskFree infrastructure audit, reply within 2 hours
What this solves
Backups nobody has ever restored
A dump lands in a folder on the same machine it came from, and no one has unpacked it since it was set up. Until a restore has been rehearsed you do not have a recovery time, you have a hope.
Customers report outages before you see them
No metrics, no alerting: a full disk or a crashed container reaches you as a support ticket. Monitoring closes that gap, so an incident starts with a notification instead of an angry email from your biggest account.
A cloud bill that keeps drifting up
Idle instances, snapshots from experiments that ended last year, a database two sizes larger than the load needs. A line-by-line review almost always finds spend that buys nothing, without touching redundancy.
Every deploy is a maintenance window
Containers restart, migrations run blind, users see 502s for ten minutes. Health-checked deploys keep the old version serving traffic until the new one answers correctly, and roll back on their own when it does not.
A full-time DevOps hire is more than you need
Most teams under twenty people need infrastructure work in bursts, then a few hours a month. Hiring for that means paying a salary for idle capacity, when a retainer from $400 a month covers the same ground.
How the work runs
Audit and plan
1–2 daysWe look at what runs where, how a release actually goes out, what happens to backups, and what the hosting bill is made of. You get a risk list in priority order and an estimate in hours.
Containerisation
3–5 daysDockerfiles and Compose for every service, secrets moved into environment variables, dev, staging and production separated. The whole environment comes up with one command on any machine.
CI/CD and zero-downtime deploys
3–7 daysA pipeline that builds, tests, migrates and ships. The new version only takes traffic after its health check passes, and rolls back to the previous one automatically if it fails.
Monitoring, alerting and restorable backups
2–4 daysHost and application metrics, logs in one place, alerts on disk, memory and error spikes. Backups go to off-host storage, and we restore from them in front of you with a stopwatch running.
Handover and ongoing support
1–2 daysYou get the access, the repository and a README that rebuilds the environment from zero. One month of warranty is included; a monthly retainer afterwards is optional, not assumed.
Stack
We fit the stack to the size of the problem: a two-service product does not need a Kubernetes cluster or the operational cost that comes with one. Complexity goes up only when traffic, team size or a security requirement genuinely calls for it.
- Docker
- Docker Compose
- GitHub Actions
- GitLab CI
- Nginx
- Caddy
- PostgreSQL 16
- Prometheus
- Grafana
- Loki
- Ansible
- S3-compatible backup storage
What it costs
Containerise and deploy
$600 – $900 / 60 000 – 90 000 ₽Dockerfiles and Compose for the app and its database, a one-command deploy, secrets pulled out of the repo, daily backups to off-host storage. Around 20–30 hours. More if there are over three services or data has to move without downtime.
Full infrastructure setup
$1,500 – $3,000 / 150 000 – 300 000 ₽CI/CD with tests and migrations, a staging environment, zero-downtime deploys with automatic rollback, metrics and alerting, backups verified by a real restore, and documentation that rebuilds it all. 50–100 hours, depending on service count and how well the current setup is written down.
Ongoing engineering
from $400 / from 40 000 ₽ per monthA dedicated engineer on retainer: alerts watched, images and dependencies patched, releases shipped, a restore drill every quarter. Hours are agreed up front, and anything past them is billed from $30 an hour.
Prices are indicative and do not constitute a public offer. The final cost, scope and timeline are fixed in a contract or invoice after we have discussed the task.
Related case study
The clearest example is the CRM we built from scratch for a large HoReCa service company. Twelve-plus modules and 89 database migrations only stay manageable when migrations run inside the deploy pipeline rather than by hand on the server. With RBAC covering 9 roles and 87 permissions, a bad release touches every user at once, so deploys there are health-checked with rollback.
Read the KMD caseFrequently asked questions
- How much does a DevOps setup cost?
- $600–$900 to containerise an application and give it a repeatable deploy, and $1,500–$3,000 for the full setup with CI/CD, staging, monitoring and restorable backups. Everything bills from $30 an hour, so an estimate is a number of hours you can check rather than an opaque package. We quote after a free audit, usually the next day, on 50% upfront and 50% on acceptance.
- How long does the work take?
- Nine to eighteen working days for a typical stack of a web app, a database and a couple of background workers. Containerisation is 3–5 days, CI/CD with zero-downtime deploys 3–7 days, monitoring and backups another 2–4 days. The cutover itself is scheduled for your quietest window, and downtime is usually limited to a DNS switch and a few minutes of database movement.
- Do we need Kubernetes?
- Almost certainly not. Docker Compose on one or two servers covers what most products need and costs far less to operate, because any engineer can read it, not only the one who set it up. We propose Kubernetes when you have many services, spiky load and someone in-house who will run it day to day — otherwise it is complexity you pay for twice.
- How do you know the backups actually work?
- We restore from them in front of you and time it. A backup nobody has unpacked is just a file, so we set up off-host storage rather than a folder on the same server, scheduled rotation, and a test restore into a separate environment. After that you know your real recovery time in hours and exactly what a worst-case restore loses.
- Can you reduce our cloud bill?
- Usually yes, and we start there because the savings often pay for the engagement. We go through the invoice line by line: instances nobody turned off, volumes and snapshots left from old experiments, a managed database sized for load that never arrived, egress that a cache would absorb. We only cut what does not affect reliability — backups and redundancy stay.
- Do we have to move off our current provider?
- No, if what you have is healthy we work with it. Most projects stay on the same AWS, Hetzner or DigitalOcean account and only the way they are operated changes: Docker, a pipeline, monitoring and backups appear. We suggest a migration only when the platform itself is the problem — no snapshots, no private networking, or a price well above market for the same specs.
- Is a retainer cheaper than hiring a DevOps engineer?
- For most teams under twenty people, yes, because infrastructure work is bursty. A retainer starts at $400 a month for a dedicated engineer who already knows your system, versus a full-time salary that mostly buys idle capacity. If your needs grow past that, we will say so directly and help you brief the hire instead of stretching the retainer.
- What do we own at the end?
- All of it. Source code, infrastructure definitions, pipelines and documentation transfer to you once the project is paid in full, on your own accounts and repositories rather than ours. Delivered functionality carries a one-month warranty, and we invoice in USD — USDT on TRC-20 or ERC-20 is our preferred rail, with bank transfer also fine.
Start with a free infrastructure audit
Send access, or just describe how a release goes out today. You get a prioritised list of risks and an estimate in hours, with a reply within 2 hours and no obligation.
Discuss your task