Services overview
Šis turinys dar neišverstas į lietuvių kalbą.
Debesis offers eight services. That is deliberate: a curated set that is actually operated well beats a catalogue of two hundred that nobody on your team has time to evaluate.
Compute
Section titled “Compute”Runs your container image. This is where your application lives.
Supports a listening port, external exposure, health probes, CPU and memory limits, optional autoscaling between a minimum and maximum replica count, mounted volumes, and environment variables and secrets. Command and arguments can override the image entrypoint.
Managed PostgreSQL and MySQL, with automated backups and a built-in proxy for safe access.
| Tier | Engine | Topology |
|---|---|---|
| PostgreSQL 16 — single instance | PostgreSQL 16 | One server with its own storage volume |
| PostgreSQL 17 — high availability | PostgreSQL 17 | Three servers on separate machines |
| MySQL 8 — single instance | MySQL 8 | One server with its own storage volume |
| MySQL 26 — high availability | MySQL 26 | Three servers on separate machines |
A single-instance tier tolerates a short restart while the server moves to another machine. Both engines also have a high-availability tier of three servers on separate machines, and in both cases your application uses one endpoint that always points at the server currently accepting writes:
- PostgreSQL 17 HA keeps the three in step by streaming replication with automatic failover.
- MySQL 26 HA uses MySQL’s own group replication, single-primary, with automatic election of the next primary.
Document Database
Section titled “Document Database”A managed, MongoDB-compatible document database. Existing MongoDB drivers, connection
strings and tools connect to it without a rewrite — the connection URI scheme is mongodb.
| Tier | Shape | Topology |
|---|---|---|
| Document database — single node | 0.5 vCPU / 2 GB / 25 GB, or 1 vCPU / 4 GB / 50 GB | One server |
| Document database — high availability | the same two shapes, per server | Three servers with automatic failover |
Search
Section titled “Search”Managed OpenSearch 3 — full-text search, aggregations and vector search.
OpenSearch is open source under the Apache 2.0 licence and governed by the OpenSearch Software Foundation, which is why it is the engine here: the terms you build on cannot be changed under you later.
| Tier | Shape | Topology |
|---|---|---|
| OpenSearch 3 — single node | 1 vCPU / 2 GB / 20 GB | One node, no replication |
| OpenSearch 3 — high availability | 1 vCPU / 2 GB / 20 GB per node | Three nodes, every shard replicated across separate machines |
Storage
Section titled “Storage”Block volumes and snapshots, mounted into a Compute service at a path you choose.
Messaging
Section titled “Messaging”Durable, replayable logs with consumer groups, on two engines: Apache Kafka 4 and Redis Streams. Both have a replicated tier.
- Apache Kafka 4 — a partitioned, replayable log. The single-broker tier stores each message once, so it suits development and staging. The replicated tier runs three brokers on separate machines with every partition stored three times, and acknowledges a write only once a second copy exists, so losing one machine costs no messages. Clients authenticate with SCRAM-SHA-512.
- Redis Streams (Redis 8) — the same log-and-consumer-group shape at a smaller footprint. The single-node tier also stores messages once. The replicated tier is one primary and two replicas on separate machines with automatic failover, reached through a single endpoint that follows the primary; it refuses writes rather than losing them when no replica is in step.
Managed Redis 8, for caching and ephemeral state.
The single-node tiers are one server with its own storage volume, tolerating a short restart while it moves to another machine. The replicated tier is a primary with two replicas on separate machines and automatic failover, for when an emptied cache would cost more than the tier does.
Container Registry
Section titled “Container Registry”A private image registry, so your images have somewhere to live that the platform can pull from without routing through a third party.
Platform features, not services
Section titled “Platform features, not services”Available to every tenant, and not billed as separate services:
- Secrets — sensitive configuration, injected at runtime and encrypted at rest. You can supply your own encryption key, in which case nobody at Debesis can decrypt them: not support, not an on-call engineer, not the provider. Cleartext exists only in transit, when a deploy delivers it to your cluster. If you would rather not manage a key, Debesis holds one for you — the data is still encrypted at rest, but in that case the provider can decrypt it.
- Logs — collected per service.
- Metrics — per service.
- Billing — usage and invoices, broken down per project.
What is not offered
Section titled “What is not offered”Being explicit, because the gap is worth knowing before you plan:
- No object storage. See the Storage note above.
- No managed Kubernetes. Debesis runs your containers; it does not hand you a cluster to operate.