Configuration for autoscaling

You can configure the parameters that are required for autoscaling in the rseapi.env file. The configuration is shared by the primary server and overflow servers. For the changes to take effect, restart the server.

Enable autoscaling

The autoscaling feature is disabled by default. To enable it, set these two parameters to true:

RSEAPI_REDIRECT_ENABLE=true # Default to false
RSEAPI_OVERFLOW_ENABLE=true # Default to false

Enable validating available resource

To validate available resource when handling incoming requests at a server, set RSEAPI_REDIRECT_ENABLE to true. When enabled, the server checks against its threshold setting when it accepts a new request. If the threshold is reached, the request will be forwarded to an overflow server if it is configured; otherwise, it will be denied.

Enable forwarding requests

To enable a server to forward a request to an overflow server in the overflow autoscaling mechanism, set RSEAPI_OVERFLOW_ENABLE to true. RSEAPI_REDIRECT_ENABLE=true is a prerequisite for enabling RSEAPI_OVERFLOW_ENABLE. You must set RSEAPI_REDIRECT_ENABLE=true for RSEAPI_OVERFLOW_ENABLE to be effective.

Specify the limit threshold

To specify the expected capacity of a server instance including user session and thread count, which is based on system and process configuration, set the following parameters:

#RSEAPI_MAX_USERS=50 # Default to 50
#RSEAPI_MAX_THREADS=1200 # Default to 1200

Define the overflow chain of servers

For a server defined by its server number n, call it server n, use the parameter RSEAPI_OVERFLOW_FOR_n to specify the server number of its direct overflow server. For example, RSEAPI_OVERFLOW_FOR_n=m denotes server m is the direct overflow server of server n.

Notes:

Server n is defined with RSEAPI_HTTP_PORT_n=<port_number>, and SERVER_NUM=n, or SRVNUM used in the server started task.
The port of all overflow servers must be defined per their server number.
To assign a specific JMON port for server n, use the corresponding parameter RSEAPI_JMON_PORT_n. For example, RSEAPI_JMON_PORT_n=3800. If it is not set, the default JMON port from the base configuration will be used for this server.

The defined direct overflow server can have its own direct overflow server. Continue to define its direct overflow to form the overflow chain until the last overflow server.

For example, the following chain of servers is denoted by their server numbers, with its primary server at start:

Primary 1 (P1) -> 2 -> 3 -> 4
RSEAPI_OVERFLOW_FOR_1=2
RSEAPI_OVERFLOW_FOR_2=3
RSEAPI_OVERFLOW_FOR_3=4

The port definition of the servers by their server number, for example, for server 1 and 2:

RSEAPI_PORT_HTTP_1="4082"
RSEAPI_PORT_SHUTDOWN_1="4083"
RSEAPI_PORT_HTTP_2="4091"
RSEAPI_PORT_SHUTDOWN_2="4092"

Other optional configurations

It is optional to check thread count or memory usage. To skip the check, set the following parameters. They default to false if not set.
```
#RSEAPI_SKIP_THREADS_CHECK=true
#RSEAPI_SKIP_MEMORY_CHECK=true 
```
Initially, when checking the port status of its overflow server, if the overflow port is used by another application, the server turns off the autoscaling feature. To enable a recheck of overflow server status, set the following parameter to true. It defaults to false if not set.
```
#RSEAPI_OVERFLOW_PORT_IN_USE_RECHECK=true
```
After restarting an overflow server, the server will wait for some time before checking the overflow server for its readiness to accept requests. Use this parameter to specify the time in millisecond that the server should wait before starting the check. The default is 3000 milliseconds.
```
#RSEAPI_OVERFLOW_SPAWN_WAIT_MSEC=3000
```
When a server estimates that it needs its overflow server, for example when its resource usage reaches a certain threshold level, it will check the status of its overflow server. The check can be done by placing a status checking request to the overflow server. If the direct overflow server is not up, the server will start it up.
The active status of the overflow server is cached. When the cache is expired, the server will issue the status checking request to the overflow server for a live update. Use the following parameter to set the expiration time for the cache. The default is 2 minutes.
```
#RSEAPI_OVERFLOW_CHECK_MINUTE=2
```