The existing loadbalancer can detect, and does stop sending requests to hard failed app servers.
Unf it can take up to 60 secs to detect and fail; and
The app servers frequently fail soft: Dead, but still responding to HTTP requests.
Being able to rotate logs without restarting the app servers would solve a similar problem across all of lpnet and edge; where with every rotation, we will fail around 1/4 to 1/5 of all requests until the LB detects the given apps are down.
The edge updates are simply more noticable as all 4 edge servers are updated in a single block. Taking around 15 minutes from start to finish.
The existing loadbalancer can detect, and does stop sending requests to hard failed app servers.
Unf it can take up to 60 secs to detect and fail; and
The app servers frequently fail soft: Dead, but still responding to HTTP requests.
Being able to rotate logs without restarting the app servers would solve a similar problem across all of lpnet and edge; where with every rotation, we will fail around 1/4 to 1/5 of all requests until the LB detects the given apps are down.
The edge updates are simply more noticable as all 4 edge servers are updated in a single block. Taking around 15 minutes from start to finish.