Hello YuanSheng, I also prefer the second way, as I do some analysis about the shared dict implemtation in ngx_lua module，scanning the shared dict memory is feasible.
But doing this straightforwardly may have some concurrent problem，when the healthchecker in other worker processes are writing the shared dict(adding a target for example), how do we handle this problem? Do you have any suggestions? On Mon, Mar 2, 2020 at 4:33 PM YuanSheng Wang <[email protected]> wrote: > I think the second way is better. > It's just that we need to study how it is stored in share memory. > > On Mon, Mar 2, 2020 at 10:47 AM liyong <[email protected]> wrote: > > > Hello community, > > > > If I want to know the health status for all probed targets when build an > > upstream healthy dashboard, how to effectively fetch or expose the > realtime > > or near realtime data? > > > > 1. `apisix` use the healthcheck status by calling `get_target_status` in > > `apisix.balancer.fetch_health_nodes` function at balancer_by_lua phase. > we > > may add probe here but this is a passively method, when the target does > not > > be routed, we don't know its status. > > > > 2. `lua-resty-healthcheck` use `ngx.shared.DICT` entry for each probed > > target internal health(healthy, unhealthy, mostly_healthy, > > mostly_unhealthy), do we have some way to explore the internal for > > `ngx.shared.DICT` > > > > 3. Use log collection: when the `lua-resty-healthcheck` lib have changed > > the internal state of the target, output a formatted log and we collected > > it. > > > > One more ... > > > > > -- > > *MembPhis* > My github: https://github.com/membphis > Apache APISIX: https://github.com/apache/incubator-apisix >
