Mailing-List: contact users-help@cloudstack.apache.org; run by ezmlm
Precedence: bulk
Reply-To: users@cloudstack.apache.org
Date: Wed, 22 Oct 2014 17:40:55 +0100 (BST)
From: Andrei Mikhailovsky <andrei@arhont.com>
To: users@cloudstack.apache.org
Message-ID: <25652070.9532.1413996052278.JavaMail.andrei@tuchka>
In-Reply-To: <1025288215.1383666.1413989630474.JavaMail.zimbra@saao.ac.za>
References: <1025288215.1383666.1413989630474.JavaMail.zimbra@saao.ac.za>
Subject: Re: HA issue and Xen resets
MIME-Version: 1.0
Content-Type: multipart/alternative;
	boundary="----=_Part_9531_670338.1413996052278"
Thread-Topic: HA issue and Xen resets
Thread-Index: bzoyzWxMstpcTqUutH4S5OogK/NWsp7x3vqJ

------=_Part_9531_670338.1413996052278
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit

Garith, 

Have you used HA system offering for the vms that were running on the disconnected server? you need to make sure they are HA enabled. 

I've had a play a few months ago with XenServer 6.2 and ACS 4.2.1 and HA worked perfectly well. I've disconnected the network (including storage network) from the host and after about 30 seconds the host was shown in Alert state. Shortly after i've seen the management server complaining that the guest vm has stopped unexpectidly and it was automatically restarted on another xenserver. 

What I haven't checked is what happened to the disconnected server. I have been side tracked and never checked. I think the default behaviour of the XenServer should be the host server restart. This makes sure that all vms are stopped. Can someone confirm this please? 

thanks 

Andrei 
----- Original Message -----

> From: "Garith Dugmore" <garith@saao.ac.za>
> To: users@cloudstack.apache.org
> Sent: Wednesday, 22 October, 2014 3:53:50 PM
> Subject: HA issue and Xen resets

> Hi All,

> I'm new to cloudstack and busy testing out ACS 4.3.1 on Centos 6.4
> using Xenserver 6.2. I have the management server setup and have 2
> xen servers that I'm testing out at the moment; specifically the HA
> functionality.

> After getting an instance up and running I yanked the network cable
> out of the one xen server and awaited the HA awesomeness to kick in.
> Both hosts still remained in "Up" state even though the one was no
> longer pingable and the instance that was hosted on that xen host
> still showed "Running" even though I also couldn't ping it. After
> some reading one suggestion was setting 'alert.wait' to 30 and
> restarting cloudstack-management. After that didn't seem to do
> anything after waiting a while I rebooted the management server all
> together and found that both hosts were marked as disconnected.

> I've tried going in and out of maintenance mode and ended up deleting
> the one xen host that was still reachable thinking I could just
> re-add it but I received an error in doing so. I have read somewhere
> that once you've either reinstalled the management server or removed
> a xen host you need to re-install the xen host. Is this true? I was
> hoping for a factory reset command of some sort.

> Besides my obvious HA problems and host disconnect issues which I'd
> love some pointers on are there any pointers on Xen server resets?

> Note I have attempted a 4.4.0 install on centos and after a couple
> issues that I can no longer recall I ended up with a way easier
> install on 4.3.1 which is why I've stuck with it for now.

> Any pointers will be greatly appreciated. Willing to try anything!
> --

> Garith Dugmore
> South African Astronomical Observatory
> and Southern African Large Telescope

------=_Part_9531_670338.1413996052278--