Split node healthcheck into 3: startup, readiness and liveness

[dudizimber]

Startup: returns OK - checks if the healthcheck program is running.
Liveness: returns OK if redis server returns any response, including PONG, BUSY, LOADING, etc.
Readiness: returns OK if:

• standalone: returns PONG
• replication: master -> PONG, replica -> PONG if it returns error during master synchronization
• cluster: PONG and cluster status is OK

[MuhammadQadora]

To solve this, Omnistrate has to separate the Liveness probe from the Readiness probe. Will reopen the issue once they do.

[dudizimber]

@MuhammadQadora
Following the issue where replicas sometimes use the wrong master IP address, we might want to check the master_link_status:down flag for the health of cluster and replication replicas.
What do you think?

[MuhammadQadora]

@dudizimber agree, can't think of a downside to this.

[dudizimber]

Let's also experiment using ConfigMaps to set the healthcheck code, so it can be easily changed if needed.