You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@cloudstack.apache.org by kaka chen <ka...@gmail.com> on 2012/05/21 09:47:03 UTC
Questions about cluster and HA-Mode.
Hi All:
Now we are testing the cloudstack 3.0.1 in clusted management server
environment, but we found on the all management servers will add host on
the same servers.
>From the source code, I found when KvmServerDiscoverer find the kvm
resource, it will ssh to the host, then executes a cloud-setup-agent
script, but the parameter -m always get from configuration table in the db.
So I am very confused, should we change it in the global settings
every-time, or provide a virtual IP to load-balance it when agent connect
to the management server? Thanks!
Another question: in the ha-mode document, it said a management server
determined that its peer has gone down and is unwinding any VM that the
peer was trying to start. How the source code implemented? I didn't find
the agent will reconnect to other alive management server from dead
management server. Thanks!
kaka chen
>From VLIS Lab Zhejiang University.
RE: Questions about cluster and HA-Mode.
Posted by kaka chen <ka...@gmail.com>.
Got it, thanks, Edision!
在 2012-5-22 上午1:14,"Edison Su" <Ed...@citrix.com>写道:
>
> > -----Original Message-----
> > From: kaka chen [mailto:kaka11.chen@gmail.com]
> > Sent: Monday, May 21, 2012 12:47 AM
> > To: cloudstack-dev@incubator.apache.org
> > Subject: Questions about cluster and HA-Mode.
> >
> > Hi All:
> >
> > Now we are testing the cloudstack 3.0.1 in clusted management server
> > environment, but we found on the all management servers will add host
> > on
> > the same servers.
> > From the source code, I found when KvmServerDiscoverer find the kvm
> > resource, it will ssh to the host, then executes a cloud-setup-agent
> > script, but the parameter -m always get from configuration table in the
> > db.
> > So I am very confused, should we change it in the global settings
> > every-time, or provide a virtual IP to load-balance it when agent
> > connect
> > to the management server? Thanks!
>
>
> Yes, put the management servers behind a Load-balancer, such as Netscaler:)
>
> >
> > Another question: in the ha-mode document, it said a management server
> > determined that its peer has gone down and is unwinding any VM that the
> > peer was trying to start. How the source code implemented? I didn't
> > find
> > the agent will reconnect to other alive management server from dead
> > management server. Thanks!
>
> Again, it's only happened if agent connects to a load-balancer ip address,
> not directly connects to one mgt server, as agent will automatically
> re-connect to the "host" ip address, if the connection is broken.
>
> >
> > kaka chen
> > From VLIS Lab Zhejiang University.
>
RE: Questions about cluster and HA-Mode.
Posted by Edison Su <Ed...@citrix.com>.
> -----Original Message-----
> From: kaka chen [mailto:kaka11.chen@gmail.com]
> Sent: Monday, May 21, 2012 12:47 AM
> To: cloudstack-dev@incubator.apache.org
> Subject: Questions about cluster and HA-Mode.
>
> Hi All:
>
> Now we are testing the cloudstack 3.0.1 in clusted management server
> environment, but we found on the all management servers will add host
> on
> the same servers.
> From the source code, I found when KvmServerDiscoverer find the kvm
> resource, it will ssh to the host, then executes a cloud-setup-agent
> script, but the parameter -m always get from configuration table in the
> db.
> So I am very confused, should we change it in the global settings
> every-time, or provide a virtual IP to load-balance it when agent
> connect
> to the management server? Thanks!
Yes, put the management servers behind a Load-balancer, such as Netscaler:)
>
> Another question: in the ha-mode document, it said a management server
> determined that its peer has gone down and is unwinding any VM that the
> peer was trying to start. How the source code implemented? I didn't
> find
> the agent will reconnect to other alive management server from dead
> management server. Thanks!
Again, it's only happened if agent connects to a load-balancer ip address, not directly connects to one mgt server, as agent will automatically re-connect to the "host" ip address, if the connection is broken.
>
> kaka chen
> From VLIS Lab Zhejiang University.