Hi guys,
I've done more testing around this and I can now confirm it has nothing to do
with cloudstack code.
I've tested it with rc3, reverted UEFI PR and 4.13.1 (which does not happen to
have the feature at all). Also I've used a matrix of VMware version of 6.0u2,
6.5u2 and 6.7u3.
The bug is reproducible with all the cloudstack versions, and only vmware
6.7u3, I was not able to reproduce this with 6.5/6.0. All of my results during
testing show it must be related to that specific version of VMware.
Therefore I'm reversing my '-1' and giving a +1 vote on the RC. I think it
needs to be included in release notes to refrain from that version for now
until further investigation is done.
Thanks,
Bobby.
On 19.05.20, 10:08, "Boris Stoyanov" <boris.stoya...@shapeblue.com> wrote:
Indeed it is severe, but please note it's a corner case which was unearthed
almost by accident. It falls down to using a new feature of selecting a boot
protocol and the template must be corrupted. So with already existing templates
I would not expect to encounter it.
As for recovery, we've managed to recover vCenter and Cloudstack after
reboots of the vCenter machine and the Cloudstack management service. There's
no exact points to recover for now, but restart seems to work.
By graceful failure I mean, cloudstack erroring out the deployment and VM
finished in ERROR state, meanwhile connection and operability with vCenter
cluster remains the same.
We're currently exploring options to fix this, one could be to disable the
feature for VMWare and work to introduce more sustainable fix in next release.
Other is to look for more guarding code when installing a template, since
VMware doesn’t actually allow you install that particular template but
cloudstack does. We'll keep you posted.
Thanks,
Bobby.
On 18.05.20, 23:01, "Marcus" <shadow...@gmail.com> wrote:
The issue sounds severe enough that a release note probably won't
suffice -
unless there's a documented way to recover we'd never want to leave a
system susceptible to being unrecoverable, even if it's rarely
triggered.
What's involved in "failing gracefully"? Is this a small fix, or an
overhaul? Perhaps the new feature could be disabled for VMware, or
disabled altogether until a fix is made in a patch release.
Does it only affect new templates, or is there a risk that an existing
template out in vSphere could suddenly cause problems?
On Mon, May 18, 2020 at 12:49 AM Boris Stoyanov <
boris.stoya...@shapeblue.com> wrote:
> Hi guys,
>
> A little further info on this, it appears when we use a corrupted
template
> and UEFI/Legacy mode when deploy a VM, it breaks the connection
between
> cloudstack and vCenter.
>
> All hosts become unreachable and basically the cluster is not
functional,
> have not investigated a way to recover this but seems like a huge
mess..
> Please note that user is not able to register such template in vCenter
> directly, but cloudstack allows using it.
>
> Open to discuss if we'll fix this, since it's expected users to use
> working templates, I think we should be failing gracefully and such
action
> should not be able to create downtime on such a large scale.
>
> I believe the boot type feature is new one and it's not available in
older
> releases, so this issue should be limited to 4.14/current master.
>
> Thanks,
> Bobby.
>
> On 15.05.20, 17:07, "Boris Stoyanov" <boris.stoya...@shapeblue.com>
> wrote:
>
> I'll have to -1 RC3, we've discovered details about an issue
which is
> causing severe consequences with a particular hypervisor in the
afternoon.
> We'll need more time to investigate before disclosing.
>
> Bobby.
>
> On 15.05.20, 9:12, "Boris Stoyanov" <boris.stoya...@shapeblue.com>
> wrote:
>
> +1 (binding)
>
> I've executed upgrade tests with the following configurations:
>
> 4.13.1 with KVM on CentOS7 hosts
> 4.13 with VMware6.5 hosts
> 4.11.3 with KVM on CentOS7 hosts
> 4.11.2 with XenServer7 hosts
> 4.11.1 with VMware 6.7
> 4.9.3 with XenServer 7 hosts
> 4.9.2 with KVM on CentOS 7 hosts
>
> Also I've run basic lifecycle operations on the following
> components:
> VMs
> Volumes
> Infra (zones, pod, clusters, hosts)
> Networks
> and more
>
> I did not come across any problems during this testing.
>
> Thanks,
> Bobby.
>
>
> On 11.05.20, 18:21, "Andrija Panic" <andrija.pa...@gmail.com>
> wrote:
>
> Hi All,
>
> I've created a 4.14.0.0 release (RC3), with the following
> artefacts up for
> testing and a vote:
>
> Git Branch and Commit SH:
>
>
https://gitbox.apache.org/repos/asf?p=cloudstack.git;a=shortlog;h=refs/heads/4.14.0.0-RC20200511T1503
> Commit: 6f96b3b2b391a9b7d085f76bcafa3989d9832b4e
>
> Source release (checksums and signatures are available at
the
> same
> location):
>
https://dist.apache.org/repos/dist/dev/cloudstack/4.14.0.0/
>
> PGP release keys (signed using 3DC01AE8):
> https://dist.apache.org/repos/dist/release/cloudstack/KEYS
>
> The vote will be open until 14th May 2020, 17.00 CET
(72h).
>
> For sanity in tallying the vote, can PMC members please be
> sure to indicate
> "(binding)" with their vote?
>
> [ ] +1 approve
> [ ] +0 no opinion
> [ ] -1 disapprove (and reason why)
>
> Additional information:
>
> For users' convenience, I've built packages from
> 6f96b3b2b391a9b7d085f76bcafa3989d9832b4e and published RC3
> repository here:
> http://packages.shapeblue.com/testing/41400rc3/ (CentOS
7 and
> Debian/generic, both with noredist support)
> and here
>
>
https://download.cloudstack.org/testing/4.14.0.0-RC20200506T2028/ubuntu/bionic/
> (Ubuntu 18.04 specific, no noredist support - thanks to
> Gabriel):
>
> The release notes are still work-in-progress, but for the
> upgrade
> instructions (including the new systemVM templates) you
may
> refer to the
> following URL:
>
>
https://acs-www.shapeblue.com/docs/WIP-PROOFING/pr112/upgrading/index.html
>
> 4.14.0.0 systemVM templates are available from here:
> http://download.cloudstack.org/systemvm/4.14/
>
> NOTES on issues fixed in this RC3 release:
>
> (this one does *NOT* require a full retest if you were
testing
> RC1/RC2
> already - just if you were affected this issue):
> - https://github.com/apache/cloudstack/pull/4064 - affects
> hostnames when
> attaching a VM to additional networks
>
> Regards,
>
>
> Andrija Panić
>
>
>
>
>
> boris.stoya...@shapeblue.com
> www.shapeblue.com
> 3 London Bridge Street, 3rd floor, News Building, London SE1
9SGUK
> @shapeblue
>
>
>
>
>
>
> boris.stoya...@shapeblue.com
> www.shapeblue.com
> 3 London Bridge Street, 3rd floor, News Building, London SE1 9SGUK
> @shapeblue
>
>
>
>
boris.stoya...@shapeblue.com
www.shapeblue.com
3 London Bridge Street, 3rd floor, News Building, London SE1 9SGUK
@shapeblue
boris.stoya...@shapeblue.com
www.shapeblue.com
3 London Bridge Street, 3rd floor, News Building, London SE1 9SGUK
@shapeblue
