load balancing - Force unhealthy instance for rolling code updates? -

load balancing - Force unhealthy instance for rolling code updates? -

i'm looking google compute's load balancing , need figure out way efficiently code roll. tried making our app respond 500 status code pool still reports instance "healthy". below configuration:

http health check request-path returns "up" if our cpu/memory usage not above threshold. tried making path homecoming 500 http status code outputs "500" page right status code header.

+-------------------+-------------+------+------+------------------+ | name | description | host | port | request-path | +-------------------+-------------+------+------+------------------+ | lb-usc1-app-check | | | 8888 | /monitoring/si4g | +-------------------+-------------+------+------+------------------+

target pool (contains 6 instances)

+------------------+-------------+-------------+ | name | description | part | +------------------+-------------+-------------+ | lb-usc1-app-pool | | us-central1 | +------------------+-------------+-------------+

forwarding rules

+-----------------+-------------+-------------+---------------+----------+------------+------------------------------------------+ | name | description | part | ip | protocol | port-range | target | +-----------------+-------------+-------------+---------------+----------+------------+------------------------------------------+ | lb-usc1-app-443 | | us-central1 | 23.236.56.210 | tcp | 443-443 | us-central1/targetpools/lb-usc1-app-pool | +-----------------+-------------+-------------+---------------+----------+------------+------------------------------------------+ | lb-usc1-app-80 | | us-central1 | 23.236.56.210 | tcp | 80-80 | us-central1/targetpools/lb-usc1-app-pool | +-----------------+-------------+-------------+---------------+----------+------------+------------------------------------------+

further details however, appears if run service nginx stop instance study down. not right reply need study current instance beingness deployed unhealthy, wait connections finish, , deploy our code.

also, if matters, "session affinity" on pool set "none".

so, in short, best way forcefulness ge load balancer see instance "unhealthy"?

it appears needed set status code 503 , started working.

load-balancing google-compute-engine