[jira] [Commented] (CLOUDSTACK-6975) Service monitoring starts dnsmasq on backup router when using redundant VRs

["dsclose (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CLOUDSTACK-6975?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15201479#comment-15201479 ] 

dsclose commented on CLOUDSTACK-6975:
-------------------------------------

It's not clear whether this issue persists. What is clear is that this was a Hypervisor independent issue. We overcame it by patching the monitorServices.py script to distinguish between master and backup redundant routers (which was less than ideal.)

In the current Cloudstack 4.8.0 there are at least two relevant changes:

1. dnsmasq runs by default on the backup router.
2. the monitorServices.py script appears not to run at all - that's despite the EnableServiceMonitoring setting still being enabled.

Point 2 may just be down to the script being depricated without being removed from the System VM ISO. Point 1 is caused by the reliance on a keepalived notify script to stop dnsmasq when the node transitions to backup. The problem being that such a transition only occurs during a failover. On a new network, only the master goes through a keepalived transition. We can demonstrate this using the entries in /var/log/messages. On the master router, we see the following:

root@r-3366-VM:~# egrep 'Keepalived|keepalived' /var/log/messages
Mar 17 14:19:02 r-3366-VM cloud: Starting keepalived
Mar 17 14:19:36 r-3366-VM Keepalived_vrrp[2529]: VRRP_Instance(inside_network) Transition to MASTER STATE
Mar 17 14:19:37 r-3366-VM Keepalived_vrrp[2529]: VRRP_Instance(inside_network) Entering MASTER STATE
root@r-3366-VM:~#

On the backup router, we only see the first line:

root@r-3367-VM:~# egrep 'Keepalived|keepalived' /var/log/messages
Mar 17 14:21:25 r-3367-VM cloud: Starting keepalived
root@r-3367-VM:~#

I'm still looking into the effects of the new scripts. I'll report back if I find that the dnsmasq issue is still a problem that needs to be resolved.

> Service monitoring starts dnsmasq on backup router when using redundant VRs
> ---------------------------------------------------------------------------
>
>                 Key: CLOUDSTACK-6975
>                 URL: https://issues.apache.org/jira/browse/CLOUDSTACK-6975
>             Project: CloudStack
>          Issue Type: Bug
>      Security Level: Public(Anyone can view this level - this is the default.) 
>          Components: Virtual Router
>    Affects Versions: 4.3.0
>         Environment: KVM
>            Reporter: Magnus Bengtsson
>
> When using a network offering with redundant routers the default setting of EnableServiceMonitoring enables monit on the routers.
> "network.router.EnableServiceMonitoring	service monitoring in router enable/disable option, default true	false"
> This causes monit to run dnsmasq on both the Master and Backup router. When the backup router assigns dhcp to clients, password and userdata retrieval from that node fails since the services are not running on the backup router.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)