Skip to main content

​
Overview

Each Redis data pod runs an instance manager process that exposes an HTTP API on port 8080. The operator communicates with instance managers via pod IP directly (never through Services) for precise targeting. Base URL: http://<pod-ip>:8080 Authentication: None (internal cluster communication)

​
Health Endpoints

​
GET /healthz

Liveness probe endpoint with runtime primary isolation detection. Response:
  • 200 OK - Process is alive and healthy
  • 503 Service Unavailable - Process not running or primary is isolated
Primary Isolation Logic: For primary pods with spec.primaryIsolation.enabled=true:
  1. Check if redis-server process is running
  2. Query Redis INFO replication to confirm role is master
  3. Attempt to reach Kubernetes API server (timeout: spec.primaryIsolation.apiServerTimeout)
  4. If API unreachable, attempt to reach at least one peer instance manager (timeout: spec.primaryIsolation.peerTimeout)
  5. Return 503 if both API and all peers are unreachable (primary is isolated)
Example: 
curl http://10.244.0.5:8080/healthz
# Response: ok
Isolation response: 
curl http://10.244.0.5:8080/healthz
# Response: primary isolated: cannot reach API server or any peer
# Status: 503

​
GET /readyz

Readiness probe endpoint. Response:
  • 200 OK - Redis responds to PING
  • 503 Service Unavailable - Redis not ready
Example: 
curl http://10.244.0.5:8080/readyz
# Response: ok

​
Replication Endpoints

​
GET /v1/status

Returns current replication status for the operator’s status collection. Method: GET Response Schema:
​
role
string
required
master or slave from Redis INFO replication
​
replicationOffset
int64
required
Current replication offset (master_repl_offset for primary, slave_repl_offset for replica)
​
connectedReplicas
int
required
Number of connected replicas (primary only, otherwise 0)
​
masterLinkStatus
string
up or down (replicas only, omitted for primary)
​
connected
bool
required
Always true (if endpoint responds)
Example Request: 
curl http://10.244.0.5:8080/v1/status
Example Response (Primary): 
{
  "role": "master",
  "replicationOffset": 123456789,
  "connectedReplicas": 2,
  "connected": true
}
Example Response (Replica): 
{
  "role": "slave",
  "replicationOffset": 123456780,
  "connectedReplicas": 0,
  "masterLinkStatus": "up",
  "connected": true
}

​
POST /v1/promote

Promotes a replica to primary by issuing REPLICAOF NO ONE. Method: POST Request Body: None Response:
  • 200 OK - Promotion succeeded
  • 500 Internal Server Error - Promotion failed
Example: 
curl -X POST http://10.244.0.6:8080/v1/promote
# Response: promoted

​
POST /v1/demote

Demotes an instance to replica by issuing REPLICAOF <primaryIP> <port>. Method: POST Query Parameters:
​
primaryIP
string
required
IP address of the primary instance
​
port
int
default:"6379"
Port of the primary instance
Response:
  • 200 OK - Demotion succeeded
  • 400 Bad Request - Missing or invalid parameters
  • 500 Internal Server Error - Demotion failed
Example: 
curl -X POST "http://10.244.0.7:8080/v1/demote?primaryIP=10.244.0.5&port=6379"
# Response: demoted

​
Backup Endpoint

​
POST /v1/backup

Executes a backup and uploads to object storage. Method: POST Request Schema:
​
backupName
string
required
Name of the RedisBackup resource (used for object key generation)
​
method
BackupMethod
default:"rdb"
Backup methodEnum values:
  • rdb - BGSAVE
  • aof - BGREWRITEAOF
​
destination
BackupDestination
required
Object storage destination
Response Schema:
​
artifactType
BackupArtifactType
required
Format of uploaded artifactValues:
  • rdb - RDB file
  • aof-archive - AOF directory as .tar.gz
​
backupPath
string
required
Full S3 URI of uploaded backupExample: s3://my-bucket/backups/my-backup.rdb
​
objectKey
string
required
Object key within bucketExample: backups/my-backup.rdb
​
backupSize
int64
required
Size of uploaded artifact in bytes
​
checksumSHA256
string
SHA256 checksum of uploaded artifact
Example Request: 
curl -X POST http://10.244.0.6:8080/v1/backup \
  -H "Content-Type: application/json" \
  -d '{
    "backupName": "my-backup",
    "method": "rdb",
    "destination": {
      "s3": {
        "bucket": "my-backup-bucket",
        "path": "redis-backups",
        "region": "us-west-2"
      }
    }
  }'
Example Response: 
{
  "artifactType": "rdb",
  "backupPath": "s3://my-backup-bucket/redis-backups/my-backup.rdb",
  "objectKey": "redis-backups/my-backup.rdb",
  "backupSize": 1073741824,
  "checksumSHA256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
}
Workflow:

​
RDB Method

  1. Issue BGSAVE to Redis
  2. Poll INFO persistence every 250ms until rdb_bgsave_in_progress=0
  3. Verify rdb_last_bgsave_status=ok
  4. Read /data/dump.rdb
  5. Upload to S3 with computed SHA256
  6. Return response

​
AOF Method

  1. Issue BGREWRITEAOF to Redis
  2. Poll INFO persistence every 250ms until aof_rewrite_in_progress=0
  3. Verify aof_last_bgrewrite_status=ok
  4. Create .tar.gz archive of /data/appendonlydir/
  5. Upload to S3 with computed SHA256
  6. Delete temporary archive file
  7. Return response
Timeout: 5 minutes (300 seconds) Credentials: Loaded from /backup-credentials mount (see RedisBackup credentials)

​
Metrics Endpoint

​
GET /metrics

Prometheus metrics endpoint exposing Redis metrics scraped from INFO command. Method: GET Response Format: Prometheus text format Example: 
curl http://10.244.0.5:8080/metrics
Sample Metrics: 
# HELP redis_up Whether the Redis server is up
# TYPE redis_up gauge
redis_up{cluster="my-redis",namespace="default",pod="my-redis-0"} 1

# HELP redis_connected_replicas Number of connected replicas
# TYPE redis_connected_replicas gauge
redis_connected_replicas{cluster="my-redis",namespace="default",pod="my-redis-0"} 2

# HELP redis_replication_offset_bytes Current replication offset
# TYPE redis_replication_offset_bytes gauge
redis_replication_offset_bytes{cluster="my-redis",namespace="default",pod="my-redis-0"} 123456789
Labels:
  • cluster - RedisCluster name
  • namespace - Kubernetes namespace
  • pod - Pod name

​
Sentinel Endpoints

Sentinel pods expose a minimal HTTP API:
  • GET /healthz - Process liveness (checks redis-sentinel process)
  • GET /readyz - Redis Sentinel PING check
Sentinel pods do not expose /v1/* or /metrics endpoints.

Build docs developers (and LLMs) love

Get started for freeTalk to us