Bug: when no endpoints is healthy, "it will cause crazy retries"

[Yiyiyimu]

#109 fixed the situation when some endpoints got taints it could be jumped over. But when all endpoints are unhealthy, it keeps the old method, so the problem of #101 still not fixed and "it will cause crazy retries". See discussion

[Yiyiyimu]

ping @tzssangglass @nic-chen @membphis

[membphis]

I added a bug, let us fix it

[tzssangglass]

so we stop requesting etcd when all endpoints are unhealthy?

[nic-chen]

How about increasing the waiting time?

[tzssangglass]

How about increasing the waiting time?

ngx.sleep?

[membphis]

It is more reasonable for the caller to add processing logic.

[tzssangglass]

provide some processing strategies for caller to choose from?

• stop requesting etcd(default)
• ngx.sleep
• keeps the old method

[tzssangglass]

It is more reasonable for the caller to add processing logic.

are you saying let the caller write the code themselves and have healthcheck call it?

[membphis]

When there is no healthy node available, we can directly return an error message.
The caller decides how long to sleep based on the return error message.

[tzssangglass]

When there is no healthy node available, we can directly return an error message.
The caller decides how long to sleep based on the return error message.

agree +1

[membphis]

we can check the process way in APISIX ^_^

[tzssangglass]

assigned me, fix soon