As our load increased, we scaled up accordingly, and the current keep_alive interval, which is set to 3 seconds, is too aggressive. By increasing the interval, we allow the VmHosts to catch up with the keepalive packets. Right now, we are getting many SSH exceptions in LoadBalancerVmPort monitor like this: class IOError, message: closed stream for loadbalancer healthchecks.
5.1 KiB
5.1 KiB