cloudstack-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kirk Kosinski <kirkkosin...@gmail.com>
Subject Re: HA for VMWare
Date Fri, 12 Jul 2013 08:46:24 GMT
Unmanaging a cluster can sometimes help fix problems like a host being
"stuck" in Disconnected or Alert state.  However, the main use is for
performing cluster maintenance.  CloudStack won't do anything with an
unmanaged cluster, including deploying VMs.  Also, when a cluster is
re-managed, CloudStack will re-sync the VM locations so nothing bad will
happen to VMs moved during maintenance.

Normally VMs being moved by hypervisor management tools is not a problem
with vSphere clusters, but it is not safe with XS and KVM clusters.
Before unmanage, you needed to setup iptables or firewall rules to block
communication between the management servers and hypervisor hosts during
maintenance, and before reestablishing communications you needed to
either move all VMs back to their original host or manually update the
database with the correct host.

Best regards,
Kirk

On 07/11/2013 05:12 PM, Mike Tutkowski wrote:
> Hey Kirk,
> 
> Thanks for that info.
> 
> I was going to ask about the "unmanage" feature of a cluster that you
> brought up.
> 
> Can you give me an idea of what unmanage does (or doesn't do)? Like when
> you would use that ability.
> 
> Thanks!
> 
> 
> On Thu, Jul 11, 2013 at 4:00 PM, Kirk Kosinski <kirkkosinski@gmail.com
> <mailto:kirkkosinski@gmail.com>> wrote:
> 
>     Hi, Mike.  CloudStack does it's own HA for KVM and XCP/XenServer.  AFAIK
>     there is no native HA for KVM.  For XCP/XS, you should not enable the
>     native HA.  It is not tested or supported with CloudStack.  Unless you
>     unmanage an XCP/XS cluster in CloudStack, moving VMs around directly
>     using the hypervisor management tools (native HA, XenCenter, xe
>     commands) is not supported and can cause problems.
> 
>     Additionally, not all guest networks will be necessarily be available on
>     all hosts at all times, and the native HA might not be able to handle
>     this since it is non-default behavior.  CloudStack manages guest
>     networks on a per-host basis, rather than per-pool (the default), to
>     conserve host resources.  Overall I would recommend using the CloudStack
>     HA for KVM and XCP/XenServer clusters.
> 
>     Best regards,
>     Kirk
> 
>     On 07/10/2013 10:53 PM, Mike Tutkowski wrote:
>     > This is sort of related to this topic.
>     >
>     > It sounds like CS relies on vCenter for HA.
>     >
>     > How about for XenServer and KVM?
>     >
>     > I don't think KVM has any notion of HA, so I assume CS handles it
>     for KVM.
>     > Is that true?
>     >
>     > With XenServer, is CS the only entity that handles HA or can that be
>     > enabled within XenServer and CS is OK with that (as long as it is
>     perhaps
>     > not also enabled within CS)?
>     >
>     > Thanks!
>     >
>     >
>     > On Wed, Jul 10, 2013 at 11:22 PM, Koushik Das
>     <koushik.das@citrix.com <mailto:koushik.das@citrix.com>>wrote:
>     >
>     >> That's right. Cloudstack relies on native HA capabilities provided by
>     >> vSphere.
>     >>
>     >>> -----Original Message-----
>     >>> From: Kirk Kosinski [mailto:kirkkosinski@gmail.com
>     <mailto:kirkkosinski@gmail.com>]
>     >>> Sent: Thursday, July 11, 2013 6:30 AM
>     >>> To: users@cloudstack.apache.org <mailto:users@cloudstack.apache.org>
>     >>> Cc: Chip Childers; dev@cloudstack.apache.org
>     <mailto:dev@cloudstack.apache.org>
>     >>> Subject: Re: HA for VMWare
>     >>>
>     >>> Hi, unless something changed (I checked 4.0/4.1 release notes
>     and found
>     >>> nothing), CloudStack uses the native vSphere HA (unlike
>     XenServer and
>     >>> KVM), so HA must be enabled in a CloudStack-managed vSphere for
>     HA to
>     >>> work.  Shutting down a VM through vCenter is probably seen as a
>     valid
>     >>> shutdown (same as shutting down through CS UI/API) and doesn't
>     trigger HA
>     >>> even if it is enabled in vSphere.  This might be different for other
>     >> hypervisors
>     >>> since VMSync for vSphere is different.
>     >>>
>     >>> Best regards,
>     >>> Kirk
>     >>>
>     >>> On 07/10/2013 01:03 PM, Chip Childers wrote:
>     >>>> It "should" work for CS to do the HA (typically with the VMware
>     >>>> cluster *not* having HA enabled).
>     >>>>
>     >>>> Nicolas, perhaps open a bug?
>     >>>>
>     >>>> On Wed, Jul 10, 2013 at 4:01 PM, Jörgen Maas
>     <jorgen.maas@gmail.com <mailto:jorgen.maas@gmail.com>>
>     >>> wrote:
>     >>>>> This should be done by ESX instead of CS, to CS ESX hypervisor
is
>     >>>>> externally managed (vCenter) I guess you need to enable HA in
your
>     >>>>> vmware configuration
>     >>>>>
>     >>>>>
>     >>>>> On Wed, Jul 10, 2013 at 12:11 PM,
>     <nicolas.lamirault@orange.com <mailto:nicolas.lamirault@orange.com>>
>     >> wrote:
>     >>>>>
>     >>>>>> Hi,
>     >>>>>> We are testing CS 4.1.0 with VMWare vSphere 5.0.
>     >>>>>> If we stop a VM using vCenter, CS doesn't try to restart
it.
>     In logs
>     >>>>>> we see :
>     >>>>>>
>     >>>>>> Skip HA for VMware VM i-xxxxxx
>     >>>>>>
>     >>>>>> In source code, we can see :
>     >>>>>>
>     >>>>>> @Override
>     >>>>>>     public void scheduleRestartForVmsOnHost(final HostVO
host,
>     >>>>>> boolean
>     >>>>>> investigate) {
>     >>>>>>
>     >>>>>>  [...]
>     >>>>>>
>     >>>>>>  if(host.getHypervisorType() == HypervisorType.VMware) {
>     >>>>>>             s_logger.info <http://s_logger.info>("Don't
>     restart for VMs on host " +
>     >>>>>> host.getId() + " as the host is VMware host");
>     >>>>>>             return;
>     >>>>>>         }
>     >>>>>>
>     >>>>>>
>     >>>>>>  [...]
>     >>>>>>
>     >>>>>> }
>     >>>>>>
>     >>>>>>
>     >>>>>> So, CS does not care to restart the VM ?
>     >>>>>> Regards.
>     >>>>>>
>     >>>>>>
>     >>>>>> --
>     >>>>>> Nicolas Lamirault
>     >>>>>>
>     >>>>>>
>     >>>>>>
>     >>> __________________________________________________________
>     >>> __________
>     >>>>>> _____________________________________________________
>     >>>>>>
>     >>>>>> Ce message et ses pieces jointes peuvent contenir des
>     informations
>     >>>>>> confidentielles ou privilegiees et ne doivent donc pas etre
>     >>>>>> diffuses, exploites ou copies sans autorisation. Si vous
avez
>     recu
>     >>>>>> ce message par erreur, veuillez le signaler a l'expediteur
et le
>     >>>>>> detruire ainsi que les pieces jointes. Les messages electroniques
>     >>>>>> etant susceptibles d'alteration, Orange decline toute
>     responsabilite
>     >>>>>> si ce message a ete altere, deforme ou falsifie. Merci.
>     >>>>>>
>     >>>>>> This message and its attachments may contain confidential
or
>     >>>>>> privileged information that may be protected by law; they
>     should not
>     >>>>>> be distributed, used or copied without authorisation.
>     >>>>>> If you have received this email in error, please notify
the
>     sender
>     >>>>>> and delete this message and its attachments.
>     >>>>>> As emails may be altered, Orange is not liable for messages
that
>     >>>>>> have been modified, changed or falsified.
>     >>>>>> Thank you.
>     >>>>>>
>     >>>>>>
>     >>>>>
>     >>>>>
>     >>>>> --
>     >>>>> Grtz,
>     >>>>> Jörgen Maas
>     >>
>     >>
>     >
>     >
> 
> 
> 
> 
> -- 
> *Mike Tutkowski*
> *Senior CloudStack Developer, SolidFire Inc.*
> e: mike.tutkowski@solidfire.com <mailto:mike.tutkowski@solidfire.com>
> o: 303.746.7302
> Advancing the way the world uses the cloud
> <http://solidfire.com/solution/overview/?video=play>/™/


Mime
View raw message