Any thoughts on releasing IPv6 address from the dnsmasq without restart ?
dhcp_release man page states that it takes dotted-quad ip address.

Thanks,
Jayapal


On 04-Jul-2013, at 4:48 PM, Jayapal Reddy 
<jayapalreddy.ur...@citrix.com<mailto:jayapalreddy.ur...@citrix.com>> wrote:

Hi,

The dhcp_release is sending release request on lo (loop back) interface updated 
it to eth0.
It is suppose to send release request on the interface on which dhcp server is 
listening.

Thanks,
Jayapal

On 03-Jul-2013, at 5:22 PM, Jayapal Reddy Uradi 
<jayapalreddy.ur...@citrix.com<mailto:jayapalreddy.ur...@citrix.com>> wrote:

Guest VM not able get ip address issue got reproduced (CLOUDSTACK-3312).
This issue happened  because the dnsmasq daemon still assuming the ip with the 
old mac.

Jul  3 11:16:44 dnsmasq-dhcp[2648]: not using configured address 10.223.250.227 
because it is leased to 6:00:0e:00:00:20
Jul  3 11:16:44 dnsmasq-dhcp[2648]: DHCPDISCOVER(eth0) 06:7d:92:00:00:20 no 
address available

https://issues.apache.org/jira/browse/CLOUDSTACK-2299 fix not solved the 
problem fully.

As suggested by Hugo it is better restart the dnsmasq when ever there is match 
of  ip, mac in lease file.
Send SIGHUP when there is no match.

I am planning to add above logic into edithosts.sh to fix CLOUDSTACK-3312.

Please comment.

Thanks,
Jayapal

-----Original Message-----
From: Hugo Trippaers 
[mailto:htrippa...@schubergphilis.com<http://schubergphilis.com/>]
Sent: Thursday, 2 May 2013 5:52 PM
To: dev@cloudstack.apache.org<mailto:dev@cloudstack.apache.org>
Subject: RE: Virtual Router: DHCP and 2-second DNS outages

I changed the system to hup a while back. I was having the same problems.
At that time I never ran into the leases problem, but recently I did have
problem with it as well.

The workaround I'm testing right now is to set a restart flag is the ip or mac
was found in the leases file and do a restart instead of a HUP if that flag is 
set.
Not the neatest of solutions but it work fine. Now I only get the occasional
error because of the DNS timeout instead of a constant problem.

Dennis' solution looks a lot nicer.

Cheers,

Hugo

-----Original Message-----
From: Marcus Sorensen [mailto:shadow...@gmail.com<http://gmail.com/>]
Sent: Wednesday, May 01, 2013 10:51 PM
To: dev@cloudstack.apache.org<mailto:dev@cloudstack.apache.org>
Subject: Re: Virtual Router: DHCP and 2-second DNS outages

What is scripts/network/exdhcp/dnsmasq_edithosts.sh ?

I noticed that master is already patched to hup, but it's missing the
dhcp_release that dennis just put in a review for.


On Wed, May 1, 2013 at 1:43 PM, Marcus Sorensen
<shadow...@gmail.com<mailto:shadow...@gmail.com>>
wrote:

oops, minus my two 'echo' debugging lines in there



On Wed, May 1, 2013 at 1:41 PM, Marcus Sorensen
<shadow...@gmail.com<mailto:shadow...@gmail.com>>wrote:

I'm not sure how we'd handle older versions if they don't have the
dhcp_release util, but how about this going forward?

diff --git a/patches/systemvm/debian/config/root/edithosts.sh
b/patches/systemvm/debian/config/root/edithosts.sh
index 9f21f20..f6e09d6 100755
--- a/patches/systemvm/debian/config/root/edithosts.sh
+++ b/patches/systemvm/debian/config/root/edithosts.sh
@@ -71,6 +71,7 @@ DHCP_HOSTS=/etc/dhcphosts.txt
DHCP_OPTS=/etc/dhcpopts.txt
DHCP_LEASES=/var/lib/misc/dnsmasq.leases
HOSTS=/etc/hosts
+ROUTER_VERSION=$(cat /etc/debian_version | cut -d. -f1)

source /root/func.sh

@@ -99,6 +100,14 @@ wait_for_dnsmasq () {

logger -t cloud "edithosts: update $1 $2 $3 to hosts"

+if [ $ROUTER_VERSION -gt 6 ]
+then
+  #release previous dhcp lease if present
+  echo releasing $ipv4
+  dhcp_release lo $ipv4 $(grep $ipv4 $DHCP_LEASES | awk '{print
+$2}')
/dev/null 2>&1
+  echo released $ipv4
+fi
+
[ ! -f $DHCP_HOSTS ] && touch $DHCP_HOSTS  [ ! -f $DHCP_OPTS ] &&
touch $DHCP_OPTS  [ ! -f $DHCP_LEASES ] && touch $DHCP_LEASES @@
-200,7 +209,12 @@ fi  pid=$(pidof dnsmasq)  if [ "$pid" != "" ]
then
-  service dnsmasq restart
+  if [ $ROUTER_VERSION -gt 6 ]
+  then
+    kill -s 1 $pid
+  else
+    service dnsmasq restart
+  fi
else
  if [ $no_redundant -eq 1 ]
  then




On Wed, May 1, 2013 at 12:33 PM, Chip Childers
<chip.child...@sungard.com<mailto:chip.child...@sungard.com>
wrote:

On Wed, May 01, 2013 at 12:29:32PM -0600, Marcus Sorensen wrote:
How do we go about requesting that dnsmasq-utils be installed on
the
new
system VM?

The code to create system VMs is in the repo now.  You should be
able to just make the change and / or submit a patch suggesting
the
change.

See:

tools/appliance/definitions






Reply via email to