[Dnsmasq-discuss] SIGHUP issue

[C V]

We use dnsmasq in CloudStack [1] to serve dhcp requests to VM instances in a 
cloud environment.
In a cloud environment, VMs come and go frequently so there dnsmasq gets 
reconfigured quite often.
Specifically we update the files pointed to by dhcp-hostsfile and dhcp-optsfile 
using automation scripts.

We used to send SIGHUP to dnsmasq to reload these files and clear the cache, 
however it does not work reliably.
Unfortunately we are not able to reproduce this problem "by hand" -- testing 
kill -1 <pid> seems to work 100% of the time.
The only 1 time I have been able to reproduce by hand is by trying it on a 
long-running (many days) process in a cloud that was not being used actively.

To ensure reliability we force a full reload of dnsmasq. 

This however means a fairly significant amount of DNS downtime for the other vms 
in the network. 

Has anybody else seen this behavior? Any clues on where to look? Any 
workarounds?

[1] http://incubator.apache.org/cloudstack/