cloudstack-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kurt K <k...@ihnetworks.com>
Subject Re: Cloudstack High Availability(HA) Testing
Date Tue, 16 May 2017 02:39:52 GMT
Hi,

Thanks for the update Simon, Boris.

I will check the mentioned URLs.

-Kurt

On 05/15/2017 08:48 PM, Boris Stoyanov wrote:
> Hi Kurt,
>
> Earlier this ear we’ve worked on KVM Host-HA feature which basically offers a proper
fencing for ipmi 2.0 compliant hosts and works with VM-ha as well, so if you have a failed
host or even shut down the Host-HA framework will try to recover it and if it fail will fencing
(issue IPMI off command ) and obviously VMs will be migrated to healthy host.
> Have a look here: https://github.com/apache/cloudstack/pull/1960
> FS: https://cwiki.apache.org/confluence/display/CLOUDSTACK/Host+HA
>
> By the way it’s one of the talks at ApacheCon this week.
> Boris Stoyanov
>
> boris.stoyanov@shapeblue.com
> www.shapeblue.com
> 53 Chandos Place, Covent Garden, London  WC2N 4HSUK
> @shapeblue
>    
>   
>
> On May 15, 2017, at 8:39 AM, Simon Weller <sweller@ena.com.INVALID<mailto:sweller@ena.com.INVALID>>
wrote:
>
> Kurt,
>
>
> By default, ACS will not HA VMs due to a host failure, as ACS assumes the storage could
still be mounted (machine isn't cleanly fenced off). Deleting the host from ACS  should trigger
a failover of the VMs. There's a lot of work in progress to deal with this a lot more gracefully.
A few folks have written out-of-band applications that watch the logs via a queue (e.g. RabbitMQ)
and on host failover use an api call to remove/disable the hosts to force the VMs to be restarted
on another host.
>
>
> - Si
>
>
>
>
> ________________________________
> From: Kurt K <kurt@ihnetworks.com<mailto:kurt@ihnetworks.com>>
> Sent: Saturday, May 13, 2017 2:38 AM
> To: users@cloudstack.apache.org<mailto:users@cloudstack.apache.org>
> Subject: Cloudstack High Availability(HA) Testing
>
> Hi There,
>
> We are in the middle of some confusion regarding Cloustack High
> Availability(HA) testing. In our deployment, we have used 2 hypervisor
> kvm servers. We have created a VM with HA compute offering in
> hypervisor2 server and then we have enabled maintenance mode on
> hypervisor2. After that the VM's are online from hypervisor1 server. But
> when we bring down the hypervisor2 server forcefully, the Vm's are not
> online from hypervisor1 server. We need help/suggestions on this.
>
> We are using below stack versions on all of our servers.
>
> ====
> Cloudstack - 4.6.0
> CentOS - 7.3
> ====
>
> Also both hypervisors are in same cluster.
>
> Please shed some lights to us.
>
> -Kurt
>
>
>



Mime
View raw message