How a Veteran Ops Leader Transforms DevOps into Full‑Chain Automation

赵舜东 – China, SaltStack user‑group founder, former military automation architect, now senior ops manager, author of "SaltStack技术入门与实站" and "运维知识体系".

Preface

The nickname "赵班长" comes from his time in the armed forces where colleagues called him that.

Today's talk covers four topics:

DevOps overview

Ops knowledge system and career development

Open‑source web architecture evolution for SMEs

Full‑chain automation practices

1. DevOps Talk

DevOps 1.0 focuses on continuous delivery, illustrated by a delivery loop diagram.

DevOps 2.0 expands to an IT service supply chain, requiring knowledge of agile, lean, continuous delivery, and IT service management.

The CALMS culture is introduced as "Ops criticism and self‑criticism", a weekly democratic meeting model from the military.

The speaker advocates a provocative approach: encouraging employees to resign when dissatisfied, emphasizing that low salary often reflects personal capability rather than company issues.

Effective communication is highlighted as a crucial ops skill, illustrated by a case where building relationships with developers enabled successful automation deployment.

2. Ops Knowledge System and Career Development

Many people underestimate operations; the speaker built a knowledge system to demonstrate that ops requires a breadth of knowledge comparable to other roles.

The system maps the layers of an HTTP request from client to server, covering storage (file vs. data storage), basic services, containers, OS, and essential ops tools such as bastion hosts.

Career paths are outlined: service builder → performance‑tuned engineer → architect (service integration) → productization (CMDB, monitoring, logging, etc.).

Different business domains (e‑commerce, gaming) demand distinct ops skill sets, affecting salary and career trajectory.

3. SME Open‑Source Web Architecture Evolution

From single‑machine setups to clustered architectures, the speaker emphasizes performance optimization in the single‑machine era.

Domain‑specific considerations such as cookie handling for image servers are discussed.

Session handling strategies (sticky sessions, replication, shared sessions) are compared.

Storage choices evolve from local disks, DAS, SAN to distributed file systems like GlusterFS, MooseFS, and FASTDFS for media files.

Web caching layers (DNS cache, reverse‑proxy cache, OpCache) are explained, with a PHP OpCache case reducing CPU usage by ~40%.

Database replication strategies (MySQL master‑master, MHA, MyCAT, TiDB) are reviewed, noting pros and cons.

Disaster recovery is illustrated with a real incident where a backup site was activated during a data‑center fire.

4. Automation Ops Development

The automation journey progresses through standardization, tooling, web‑based platforms, service orientation, and finally intelligent automation.

A case study describes automating Oracle patch upgrades by consolidating cron jobs via a job platform.

The full‑chain automation stack includes:

Automated OS installation (Cobbler)

Configuration management (SaltStack)

Monitoring (Zabbix)

Continuous delivery (Jenkins)

Log collection (ELK Stack)

Infrastructure automation (OpenStack, Kubernetes)

Data pipelines feed Kafka into Elasticsearch and Hadoop for analysis, with a recommendation to separate ops and business ES clusters.

Smart scaling is demonstrated: Zabbix triggers scaling, Salt‑Cloud creates OpenStack VMs, etcd stores state, SaltStack deploys environments, Jenkins handles code deployment, and automated tests verify the rollout.

5. Future Outlook

The next step after standardization, tooling, platformization, and service‑orientation is productization, offering opportunities for seasoned ops professionals to start their own ventures.

Continuous learning and skill expansion are essential; ops professionals can move from being scapegoats to central stage performers.