What happens when a developer commits a Kubernetes manifest and forgets that a database password is sitting in plain text inside it? That single mistake has launched a thousand security reviews, and it is exactly the problem that three very different tools try to solve. External Secrets Operator, Sealed Secrets, and HashiCorp Vault all claim to fix secret sprawl in Kubernetes, yet they solve it in ways that barely resemble each other once you look at the daily workflow.
Choosing between them is not a matter of picking the “most secure” option. It is a matter of matching a tool’s workflow to how a team actually ships software. A small team pushing manifests through GitOps has different needs than a platform group issuing short-lived database credentials to hundreds of pods. This article breaks down how each tool behaves day to day, where teams trip up, and what to watch heading into 2027.
The Core Problem All Three Tools Address
Kubernetes Secrets are the native object every one of these tools ultimately interacts with. By default, a Secret is just base64-encoded data stored in etcd, which is encoding, not encryption. Anyone who can read the manifest can read the password. The real challenge is getting a secret from wherever it is safely stored into a running pod without ever exposing it in a Git repository or a CI log.
Each tool answers a slightly different question. Sealed Secrets asks: how do I safely commit a secret to Git? External Secrets Operator asks: how do I pull secrets from a central store at runtime? Vault asks: how do I become the central store and control every credential’s lifecycle? Those are three different jobs, and confusing them is the most common mistake teams make.
Sealed Secrets: Encrypt Locally, Commit Safely

Sealed Secrets, a controller plus a `kubeseal` CLI originally from Bitnami and created around 2017, takes the most Git-native approach. A developer encrypts a Secret locally into a `SealedSecret` custom resource using an asymmetric key pair. Only the in-cluster controller holds the private key that can decrypt it. The encrypted blob goes into Git, and the controller turns it back into a normal Secret inside the cluster.
The appeal is obvious: your GitOps repository stays the single source of truth, and nothing readable ever lands in version control. Argo CD or Flux can sync a SealedSecret like any other manifest.
Where the Sealed Secrets workflow breaks down
The workflow has a sharp edge that catches teams off guard. SealedSecrets are bound to the controller’s sealing key. If that key is lost, every existing SealedSecret becomes permanently undecryptable. If the key is rotated without re-encrypting existing secrets, the same thing happens. Backing up the sealing key is not optional; it is a required operational step, and plenty of teams learn this the hard way during a cluster rebuild.
Rotation is the second friction point. When a password changes, someone has to re-seal it and commit a new manifest. There is no automatic refresh from a central store. For a handful of secrets that change rarely, this is fine. For dozens of credentials that rotate on a schedule, it becomes a manual chore that quietly gets skipped.
External Secrets Operator: Pull at Runtime, Commit Only References
External Secrets Operator (ESO) flips the model. Instead of committing encrypted values, teams commit an `ExternalSecret` manifest that references a path or key in an external provider such as AWS Secrets Manager, HashiCorp Vault, GCP Secret Manager, or Azure Key Vault. A `SecretStore` or `ClusterSecretStore` resource defines how to connect to that provider. The operator then pulls the actual value at runtime and writes it into a native Kubernetes Secret.
Nothing secret touches Git. The manifest says “fetch the key named prod/db-password from this store,” and the operator does the rest. ESO was accepted as a CNCF Sandbox project in 2022, which gave it broader community visibility and adoption.
The refresh interval is the workflow’s heartbeat
ESO’s defining behavior is its `refreshInterval`. The operator re-syncs on a configurable schedule, so when a value changes in the external store, the Kubernetes Secret updates automatically without a new commit. That single feature removes the manual re-sealing step entirely.
The trade-off is a runtime dependency. If the external provider is unreachable when a pod starts and the Secret has not yet been synced, the pod may fail to launch. Teams that run ESO well treat the external store as production infrastructure, not an afterthought. They also scope access tightly, because the operator’s credentials can read every secret it is allowed to reach.
HashiCorp Vault: The Central Source of Truth
HashiCorp Vault, first released in 2015, is not a Kubernetes-specific secret tool. It is a full secrets-management system that stores secrets centrally and issues short-lived credentials generated per request through secret engines. A pod can request database credentials, and Vault generates a fresh set with a lease that expires automatically.
Kubernetes consumes Vault secrets in two main ways. The Vault Agent Injector is a mutating webhook that injects secrets into pods based on annotations. The Vault Secrets Store CSI Driver mounts secrets as volumes. Either way, Vault remains the source of truth, and Kubernetes is just a consumer.
Why short-lived credentials change the design
The leasing model is the capability neither of the other two tools replicates natively. Instead of a static password living somewhere until someone rotates it, Vault can hand out credentials that expire in minutes. If they leak, the blast radius is small. That is a fundamentally different security posture, and it is why regulated industries gravitate toward Vault.
The cost is operational weight. Vault needs its own high-availability setup, unsealing process, and policy management. It is a platform to run, not a controller to install. Teams comparing deployment tools often notice the same pattern: the more capability a system offers, the more operational surface it demands. The same tension shows up in RabbitMQ vs Kafka workflow differences, where a simpler broker and a heavier event platform serve very different operational realities.
Side-by-Side: How the Workflows Actually Differ
The clearest way to see the differences is to compare what each tool commits, where the value lives, and how rotation works.
| Dimension | Sealed Secrets | External Secrets Operator | HashiCorp Vault |
|---|---|---|---|
| What goes in Git | Encrypted ciphertext | A reference to a path or key | Nothing secret, only policies |
| Where the value lives | In the SealedSecret, in Git | In the external provider | In Vault |
| Rotation | Manual re-seal and commit | Automatic via refresh interval | Lease-based and generated per request |
| Runtime dependency | The controller’s key | The external provider | Vault itself |
| Short-lived credentials | No | No, unless backed by Vault | Yes, natively |
| Operational weight | Low | Medium | High |
| GitOps fit | Strong, ciphertext in repo | Strong, references only | Indirect, via injector or CSI |
Notice that the GitOps fit row is a workflow distinction, not a security ranking. Both Sealed Secrets and ESO work well with Argo CD and Flux, but for opposite reasons. One commits encrypted data; the other commits only pointers.
Common Mistakes Teams Make
The most frequent error is treating Sealed Secrets as a rotation solution. It protects secrets at rest in Git, but it does nothing to refresh them. Teams that assume otherwise end up with stale credentials and a manual process nobody owns.
A second mistake is running ESO without thinking about failure modes. If the external store goes down during a deploy, pods that need a freshly synced Secret may stall. Good teams monitor sync status and alert on failures rather than discovering them during an incident.
A third mistake is adopting Vault for a small project. Vault works well, but it is a platform with real operational overhead. Installing it to manage five static secrets is like renting a warehouse to store a bicycle. Many teams are better served starting with ESO or Sealed Secrets and moving to Vault when short-lived credentials become a genuine requirement.
Finally, licensing deserves attention. HashiCorp changed Vault’s license from MPL 2.0 to BUSL 1.1 in August 2023, and IBM acquired HashiCorp (announced April 2024, completed around 2025). Anyone planning a 2027 rollout should verify current licensing terms before committing.
How These Tools Fit Into a Broader Stack
Secret management rarely exists in isolation. It sits alongside deployment pipelines, data layers, and the editors and platforms developers use every day. Teams that document how their tooling fits together tend to avoid surprises later. The same discipline that helps you reason about Airtable vs Notion workflow differences applies to infrastructure choices: understand what each tool actually owns before you build on it.
For teams running databases, the secret tool and the data store interact constantly. Vault’s short-lived credentials work well with relational databases, while the choice between SQL vs NoSQL workflow differences shapes how credentials and access patterns are structured in the first place. Picking the right secret workflow early saves rework when the data layer grows.
What to Watch Heading Into 2027
No one can predict product changes specific to 2027 with confidence, so treat any firm forecast with skepticism. What is reasonable to expect is continued convergence. ESO already integrates with Vault as a provider, which means teams increasingly combine them: Vault as the source of truth, ESO as the sync layer into Kubernetes. That hybrid pattern resolves much of the either-or debate.
Sealed Secrets will likely remain the lightweight choice for teams that want encrypted manifests in Git without running extra infrastructure. Its key-management caveat will not disappear, so backups and rotation discipline stay essential.
Vault’s direction depends heavily on licensing and ownership. The BUSL change and the IBM acquisition mean teams should re-check terms rather than assume the 2015-era open model still applies. Watch how the ecosystem responds, and prefer tools whose licenses fit your organization’s risk tolerance.
A Practical Decision Framework
Rather than asking which tool is best, ask three questions. First, where should the secret value live: in Git, in a cloud provider, or in a dedicated vault? Second, how often does it rotate, and can a human keep up? Third, how much operational burden can the team realistically carry?
If the answer is “rarely changes, small team, GitOps-first,” Sealed Secrets fits. If it is “lives in a cloud provider, needs automatic refresh, moderate team,” ESO fits. If it is “dynamic credentials, strict compliance, dedicated platform team,” Vault fits. Many mature organizations end up running ESO in front of Vault, which is a legitimate destination rather than a compromise.
Conclusion
The debate between External Secrets Operator, Sealed Secrets, and Vault is really a debate about workflow, not security theater. Sealed Secrets encrypts and commits, keeping Git as the source of truth but demanding manual rotation and careful key backups. ESO pulls references at runtime, refreshing automatically but depending on an external store staying reachable. Vault becomes the source of truth itself, issuing short-lived credentials at the cost of real operational weight.
The teams that get this right do not pick the most impressive tool. They pick the one whose daily workflow matches how they already build and ship software, then document the failure modes they are accepting. Heading into 2027, the smartest move is to verify current licensing, test failure scenarios before production, and remember that combining tools is often better than forcing a single winner.
FAQ
What is the main difference between External Secrets Operator and Sealed Secrets?
Sealed Secrets commits encrypted ciphertext to Git, and only the in-cluster controller can decrypt it. External Secrets Operator commits only a reference to a path or key in an external provider and pulls the actual value at runtime. One stores the protected secret in the repo; the other stores a pointer.
Can Sealed Secrets rotate secrets automatically?
No. When a secret value changes, someone must re-seal it with the `kubeseal` CLI and commit a new manifest. There is no automatic refresh from a central store. This manual step is the biggest workflow limitation compared with ESO’s refresh interval.
Does External Secrets Operator replace Vault?
Not necessarily. ESO can use Vault as one of its providers, so the two often work together. Vault acts as the source of truth and can issue short-lived credentials generated per request, while ESO handles syncing those values into native Kubernetes Secrets on a schedule.
What happens if the Sealed Secrets sealing key is lost?
Every existing SealedSecret becomes permanently undecryptable, because they are bound to that key. Backing up the sealing key is a required operational step. Rotating the key without re-encrypting existing secrets causes the same failure.
Is Vault worth the operational overhead for a small team?
Usually not at first. Vault requires high-availability setup, unsealing, and policy management, which is heavy for a handful of static secrets. Smaller teams often start with ESO or Sealed Secrets and move to Vault when short-lived credentials become a genuine requirement.
Does Vault’s licensing affect whether teams should adopt it in 2027?
It can. Vault moved from MPL 2.0 to BUSL 1.1 in August 2023, and IBM acquired HashiCorp in a deal announced in April 2024 and completed around 2025. Teams should verify current licensing terms before committing to a long-term rollout.


