Hi Ivan, I've seen this problem to. The solution is to handle your ip address by the host youre adding static by api or gui. The router VM uses the old host entries set in /etc/hosts if you cleanup vpc network. The host file will be cleared.
Kind Regards Ben Von meinem Huawei-Mobiltelefon gesendet -------- Originalnachricht -------- Betreff: DNS in virtual router not stable Von: Ivan X Yue <ivan...@ca.ibm.com> An: users@cloudstack.apache.org Cc: Hi, all, We are using Cloudstack 4.9.2 with KVM and advance networking We are having VPC network, and need to continuously adding / removing VM. One issue we found is, from time-to-time, on the VM deployed, it gets error that the hostname of the other VM is not resolved. We dig into the VR, and it seems that every time a VM is added or removed, the dnsmasq needs to be restarted. And based on /var/log/dnsmasq.log, there is 2 second delay before dnsmasq is running again. Quite likely while restart is happening, the DNS is not working and therefore VM will not be able to resolve hostname. Dec 12 15:22:55 dnsmasq[6479]: exiting on receipt of SIGTERM Dec 12 15:22:57 dnsmasq[6747]: started, version 2.62 cachesize 150 Dec 12 15:26:36 dnsmasq[6747]: exiting on receipt of SIGTERM Dec 12 15:26:38 dnsmasq[28921]: started, version 2.62 cachesize 150 We don't see this issue in CloudStack 4.11. In 4.11, we also see dnsmasq being restarted in dnsmasq.log, but there is no delay between exiting and started. Is there anything we can do in 4.9.2 to have the DNS more stable. e.g. can dnsmasq not being restarted even new VM info is added? Or can it just reload the VM info? Thanks Ivan