…ched wrong vs-version from client

…not cached

…ged adminstratively

When healthcheck program updating the rs's weight, it should consider:

* rs's weight may change adminstratively
* rs availabilty may change due to the server is closed or crashed

Healthcheck stores the original weight when it sends down notification in order to restore
the rs's weight when it gets healthy later. The problem is original weight is syncd from
dpvs periodically and not always up-to-date, which may result in incorrect weight update in
healthcheck up notification.

Signed-off-by: ywc689 <[email protected]>

Unfortunately, the CAS solution in the last commit is not effective for healthcheck
process cannot maintain the correct previous weight value. Thus resource version is
used instead, which should be passed to the lb api when the healthcheck try to update
rs, and the api would reject the request if the resouce version is outdated.

use vs version to resync target status when notification fails

Signed-off-by: ywc689 <[email protected]>

- update config version finally in updateConfig and only when the backend config is in healthy state
- fix notification resync fail problem caused by insufficient channel size
- fix array index problem in function NewTargetFromStr

Signed-off-by: ywc689 <[email protected]>

Signed-off-by: ywc689 <[email protected]>

…g. POST|PUT /vs/${ID}/rs, PUT /vs/${ID})

Healthcheck api /vs/${ID}/rs/health would not update local cache.

GET API /vs | /vs/${ID} response dpvs-agent local cache info default
and response dpvs running service detail with query param `healthcheck=true`

Signed-off-by: ywc689 <[email protected]>

… services

…checkers

Signed-off-by: ywc689 <[email protected]>

Signed-off-by: ywc689 <[email protected]>