How Sealer Streamlines Private Cloud Delivery for Large‑Scale Enterprise Apps
This article explains how the open‑source sealer tool addresses the complexity of private, large‑scale application delivery by integrating Kubernetes cluster and business components, reducing deployment effort from weeks to days while ensuring consistency and low‑dependency operations.
Background
With the rapid growth of the Internet, cloud‑native technologies centered on containers have exploded, and Kubernetes has become the de‑facto standard for container orchestration. However, while Kubernetes solves large‑scale deployment and scheduling, it is not friendly for business‑level delivery and its own installation is complex.
Challenges of Private Delivery
In the government procurement (政采云) scenario, private delivery must handle more than 300 business components and over 20 middleware pieces across heterogeneous, often isolated networks. The main pain points are handling deployment dependencies, ensuring delivery consistency, and managing the massive manual effort required for preparation and execution.
Sealer Solution
The open‑source project sealer , initiated by Alibaba Cloud’s native application platform team and co‑built with partners, fills the gap by providing an elegant design that packages an entire Kubernetes cluster together with distributed applications. Similar to Docker, sealer treats the whole cluster as a single machine, defines the “operating system” with a Kubefile, builds a cluster image, and runs it to deliver the full stack.
Problems with Ansible
Only solves deployment steps; dependencies still need separate preparation, which is costly and impossible in isolated networks.
Adapting playbooks for each customer’s unique requirements requires extensive debugging.
The declarative language lacks power for complex control logic.
Ansible requires a runtime environment, preventing a true zero‑dependency delivery.
Community Contributions
Implemented a lite build mode that parses Helm charts, YAML definitions, and image lists to cache images without launching a cluster, reducing build cost to a single host.
Added a check feature to automatically verify the status of Kubernetes components after delivery.
Enabled custom registry configuration, moving the registry from a fixed master node to a user‑defined location.
Provided a sealer join command for adding nodes to an existing cluster.
Key Features
Cluster images produced by sealer can be pushed directly to private Docker registries such as Harbor, and later extended or rebuilt.
The built‑in registry supports multi‑source, multi‑domain proxy caching, allowing private caching of public images without changing image references, and can proxy multiple private registries simultaneously.
Implementation Results
Using sealer, the delivery workflow was re‑designed: business components, middleware, and image caching are defined in a Kubefile and built with the lite mode. This eliminated most manual dependency handling, shortened the delivery cycle from 15 person‑days to 2 person‑days, and successfully delivered a cluster with 20 GB of cached images, over 2000 GB memory and 800+ CPU cores.
Future Outlook
The successful rollout demonstrates the power of open‑source collaboration. The team will continue to contribute to the sealer community, address remaining issues, and expand its capabilities to serve more scenarios, inviting more partners to co‑build a brighter, more robust tool.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
Alibaba Cloud Native
We publish cloud-native tech news, curate in-depth content, host regular events and live streams, and share Alibaba product and user case studies. Join us to explore and share the cloud-native insights you need.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
