Managed Service Settings

Closed beta. These settings belong to Managed services, which is in closed beta and enabled per organisation on request. Contact support to have it turned on for your organisation.

You choose these settings when you set up a service. Each one is a whole number within its range, and a field left empty takes the default. Changing a running service’s settings from Managed services is not available yet, so size it for what you need from the start. Three rules hold for every service:

memory_mib is both the memory request and the limit, so the service is capped at it.
cpu_millicores is a CPU request, where 1000 is one core. CPU is not capped.
storage_gib is the size of each volume, taken from the cluster’s default StorageClass. A volume never shrinks, and it can only grow where the StorageClass allows volume expansion.

PostgreSQL

Package postgresql 1.0.0: PostgreSQL 18.6, run by the CloudNativePG operator.

Setting	Range	Default	What it changes
`instances`	1 to 3	1	One primary plus `instances - 1` streaming replicas, spread across nodes where the cluster has more than one. 1 is a single database with no failover; 3 survives the loss of one node, with automatic failover. Each instance has its own volume.
`storage_gib`	1 to 1000 GiB	10	Size of each instance’s data volume.
`memory_mib`	256 to 16384 MiB	512	Memory for each instance. PostgreSQL’s `shared_buffers` stays at its 128 MB default, so the rest serves as file cache and per-connection working memory.
`cpu_millicores`	100 to 8000	250	CPU request for each instance.

With one instance, a voluntary drain of the database’s node waits until you add instances: the operator’s PodDisruptionBudget holds it, rather than moving the only copy of the data.

Valkey

Package valkey 1.0.0: Valkey 9.1.

Setting	Range	Default	What it changes
`replicas`	0 to 2	0	Read-only replicas of the primary, served through the read-replicas address. 0 is a single primary. Replicas add read capacity and a warm copy of the data, but they do not take over writes while the primary is down.
`memory_mib`	128 to 16384 MiB	256	Memory for each pod. The cache holds at most 60% of it, then evicts the least recently used keys.
`storage_gib`	1 to 100 GiB	2	Size of each pod’s snapshot volume. It needs to hold one snapshot of the data, which is at most 60% of `memory_mib`. Snapshots let a restarted pod come back with its data.
`cpu_millicores`	50 to 4000	100	CPU request for each pod.

NATS

Package nats 1.0.0: NATS 2.15 with JetStream.

Setting	Range	Default	What it changes
`replicas`	1	1	Number of servers. This version runs a single server.
`storage_gib`	1 to 1000 GiB	10	Size of the JetStream volume. Streams may use up to 90% of it.
`memory_mib`	128 to 16384 MiB	256	Memory for the server. Its Go runtime is told to stay under 80% of it.
`cpu_millicores`	50 to 4000	100	CPU request for the server.

OpenSearch

Package opensearch 1.0.0: OpenSearch 3.9, as a single node with the security plugin on.

Setting	Range	Default	What it changes
`storage_gib`	5 to 1000 GiB	20	Size of the data volume. OpenSearch stops accepting writes when it is 95% full.
`memory_mib`	1024 to 32768 MiB	2048	Memory for the node. Half of it is the JVM heap; the rest is left to the operating system’s file cache, which OpenSearch depends on.
`cpu_millicores`	250 to 8000	500	CPU request for the node.

SeaweedFS

Package seaweedfs 1.0.0: SeaweedFS 4.48 with an S3-compatible API, as one pod on one volume.

Setting	Range	Default	What it changes
`storage_gib`	1 to 2000 GiB	20	Size of the data volume, which holds every object.
`memory_mib`	256 to 16384 MiB	512	Memory for the pod.
`cpu_millicores`	100 to 8000	250	CPU request for the pod.

VictoriaLogs

Package victoria-logs 1.0.0: VictoriaLogs 1.52, as a single server.

Setting	Range	Default	What it changes
`retention_days`	1 to 365 days	14	Days a log line is kept before it is deleted. Once the volume is 80% full, the oldest data is deleted first, so logs that arrive faster than `storage_gib` allows for can go sooner.
`storage_gib`	1 to 1000 GiB	20	Size of the log volume.
`memory_mib`	256 to 16384 MiB	512	Memory for the server.
`cpu_millicores`	100 to 8000	250	CPU request for the server.

VictoriaMetrics

Package victoria-metrics 1.0.0: VictoriaMetrics 1.153, as a single server.

Setting	Range	Default	What it changes
`retention_days`	1 to 3650 days	30	Days a sample is kept before it is deleted.
`storage_gib`	1 to 1000 GiB	20	Size of the data volume. VictoriaMetrics stops accepting writes when it is full.
`memory_mib`	256 to 16384 MiB	512	Memory for the server. VictoriaMetrics sizes its caches to 60% of it.
`cpu_millicores`	100 to 8000	250	CPU request for the server.

Health checks

Ankra decides a service’s health from these checks, read from its synced view of the service’s cluster, in the service’s namespace.

Service	Workload, ready against the plan	Kubernetes Service	Secret, with every key the connection lists
PostgreSQL	CloudNativePG pods labelled `cnpg.io/cluster=postgresql`: the primary Ready, `instances` pods Ready, and one `Bound` volume claim per instance	`postgresql-rw`	`postgresql-app`
Valkey	StatefulSet `valkey`, `1 + replicas` ready	`valkey`	`valkey-credentials`
NATS	StatefulSet `nats`, `replicas` ready	`nats`	`nats-credentials`
OpenSearch	StatefulSet `opensearch`, 1 ready	`opensearch`	`opensearch-credentials`
SeaweedFS	Deployment `seaweedfs-all-in-one`, 1 available	`seaweedfs-all-in-one`	`seaweedfs-s3-secret`
VictoriaLogs	StatefulSet `victoria-logs`, 1 ready	`victoria-logs`	`victoria-logs-credentials`
VictoriaMetrics	StatefulSet `victoria-metrics`, 1 ready	`victoria-metrics`	`victoria-metrics-credentials`

Every service is also checked for the stack that owns it: a member that failed to apply or is being removed, or a stack that is gone or being removed, reads degraded.

Value	Applies to
About 1 minute	How often each service is observed
5 minutes	How long an observation stands before it reads `unknown`
15 minutes	How recent Ankra’s view of a resource type on the cluster must be for a missing object to count as missing; older reads `unknown`
10 minutes (PostgreSQL, OpenSearch), 5 minutes (the others)	Grace window after a successful deployment, during which a failing check reads `progressing`
30 minutes	How long after its last write a stack member’s apply still counts as in progress

Readiness check

Every service has a readiness check: a short-lived pod the cluster’s agent starts in the service’s namespace, which connects to the service’s primary address as a client would. <service> stands for the service name.

Service	Image	Credentials, by reference from the service’s Secret
PostgreSQL	`postgres:16-alpine`	`password` of `postgresql-app`
Valkey	`valkey/valkey:9.1.2-alpine`	`username` and `password` of `valkey-credentials`
NATS	`natsio/nats-box:0.19.3`	`username` and `password` of `nats-credentials`
OpenSearch	`curlimages/curl:8.16.0`	`username`, `password` and `ca.crt` of `opensearch-credentials`
SeaweedFS	`curlimages/curl:8.16.0`	`admin_access_key_id` and `admin_secret_access_key` of `seaweedfs-s3-secret`
VictoriaLogs	`curlimages/curl:8.16.0`	`username` and `password` of `victoria-logs-credentials`
VictoriaMetrics	`curlimages/curl:8.16.0`	`username` and `password` of `victoria-metrics-credentials`

The images other than PostgreSQL’s are pinned by digest. What each check does is listed under Readiness. The PostgreSQL check runs psql -h postgresql-rw.<service>.svc.cluster.local -p 5432 -U app -d app -c "SELECT 1".

Property	Value
Pod label	`app.kubernetes.io/name=ankra-postgres-connection-test` for PostgreSQL, `app.kubernetes.io/name=ankra-service-canary` for the others. The pod is deleted when it finishes.
Credential values	Passed to the pod by reference. They never leave the cluster.
Interval	About every 5 minutes, measured from the last attempt
Result stands	15 minutes
Pod time limit	60 seconds
Grace window	While no check has passed yet: 10 minutes after a successful deployment for PostgreSQL and OpenSearch, 5 minutes for the other services
Agent	The checks for services other than PostgreSQL need an agent recent enough to run them. Until it is, the service reads `unverified` and `readiness_check.reason` says to upgrade the agent.

At most one check pod runs on a cluster at a time.

Set up a managed service

Where these settings are chosen.

Connect to a managed service

Addresses and credentials for each service.