XCP-ng 7.5.0 Release Candidate

olivierlambert

Nope
Yes.

stormi

This thread is having several discussions in parallel. If nobody minds, I'll move them to separate threads soon.

Update: done.

Update with HA enabled / or not upgrading master first : https://xcp-ng.org/forum/topic/265/issues-when-upgrading-with-ha-enabled
USB passthrough: https://xcp-ng.org/forum/topic/266/usb-passthrough-test-reports-in-7-5rc1

stormi

I pushed several update candidates to the RC in the updates_testing repository:

vhd-tool fixing two bugs:
- "VDI import seems broken for large VHD files" https://bugs.xenserver.org/browse/XSO-868
- "VDI export fails with VDI I/O error" https://bugs.xenserver.org/browse/XSO-873
xen security update

To test: yum update --enablerepo='xcp-ng-updates_testing*' and reboot.

What we need is to make sure they cause no regressions.

Thanks!

olivierlambert

@stormi said in XCP-ng 7.5.0 Release Candidate:

yum update --enablerepo='xcp-ng-updates_testing*'

Is this interfering with the ZFS packages or there is no issue to do it? (especially blktap)

stormi

@olivierlambert No issue to do it.

olivierlambert

Great!

christianuhlmann

@christianuhlmann said in XCP-ng 7.5.0 Release Candidate:

my first host with xcp-ng upgraded successful to 7.5 (Intel NUC) without an error.
thanks for the great job !

second server has also been updated successfully.

the third server I use pci passthroug for 2 DVB cards.
So far that was excellent under 7.4.1.

Since upgrading to 7.5 I get the following error if i want to start the vm with the passthroug pci devices:

vm.start
{
  "id": "40bcca30-d07a-db7c-5744-c134a30af5d1"
}
{
  "message": "INTERNAL_ERROR(xenopsd internal error: (Failure
  \"Error from xenguesthelper: Populate on Demand and PCI Passthrough are mutually exclusive\"))",
  "stack": "XapiError: INTERNAL_ERROR(xenopsd internal error: (Failure
  \"Error from xenguesthelper: Populate on Demand and PCI Passthrough are mutually exclusive\"))
    at wrapError (/home/node/xen-orchestra/packages/xen-api/src/index.js:197:9)
    at /home/node/xen-orchestra/packages/xen-api/src/index.js:1064:18
    at run (/home/node/xen-orchestra/node_modules/core-js/modules/es6.promise.js:75:22)
    at /home/node/xen-orchestra/node_modules/core-js/modules/es6.promise.js:92:30
    at flush (/home/node/xen-orchestra/node_modules/core-js/modules/_microtask.js:18:9)
    at _combinedTickCallback (internal/process/next_tick.js:131:7)
    at process._tickCallback (internal/process/next_tick.js:180:9)",
  "code": "INTERNAL_ERROR",
  "params": [
    "xenopsd internal error: (Failure
  \"Error from xenguesthelper: Populate on Demand and PCI Passthrough are mutually exclusive\")"
  ],
  "method": "VM.start"
}

all seem ok

# lspci
.....
02:00.0 Multimedia controller: Digital Devices GmbH Octopus DVB Adapter
......
08:00.0 Multimedia controller: Digital Devices GmbH Cine V7

# xl pci-assignable-list
0000:02:00.0
0000:08:00.0

# xe vm-param-get uuid=40bcca30-d07a-db7c-5744-c134a30af5d1 param-name=other-config
pci: 0/0000:02:00.0,0/0000:08:00.0; base_template_name: Debian Stretch 9.0; import_task: OpaqueRef:83a9253c-f3c0-448f-aba6-df5efa2c1c4a; mac_seed: 1113ba83-5709-4ca5-a292-f8b81a047db4; install-methods: cdrom,nfs,http,ftp; linux_template: true

Does anyone have any idea what that might be?
does anyone have an upgrade to 7.5 with pci passthroug on the run?

Thanks and Greetings

Christian

olivierlambert

@christianuhlmann said in XCP-ng 7.5.0 Release Candidate:

Populate on Demand

Do you know what's this? It seems incompatible to have it enable in parallel with PCI Passthrought.

edit: it seems your dynamix min/max and static max aren't set at the same value.

christianuhlmann

@olivierlambert

Hi
After a bit of googling on populate on demand I realized.
strange that it worked under 7.4.1.

if I understood correctly, it is the dynamic ram assign with min / max.

after I entered a fixed size for ram, the start worked fine.

thank you for the great work

Christian

olivierlambert

I think ballooning + PCI passthrough can cause some crashes, so that's why there is a check to avoid using both at the same time.

Bungo

Updated via YUM, all seemed to go well, except the xcp-ng tools ISO only seems to have the Linux directory and the readme.txt file on it, no windows tools.

Guests on 3 hosts same.

file appears to be "guest-tools-7.29.0-2.iso" in /opt/xensource/packages/iso

cheers.

stormi

The windows tools are not free software, we can't distribute them. There's a way to install them though: https://xcp-ng.org/2018/04/23/the-future-of-vm-tools/ (see "a better model")

Bungo

The readme doesn't say that. Might need to adjust it.

stormi

Indeed. I'll fix it.

stormi

@bungo I uploaded a fixed package, can you test it?

# only needed if you test so quickly that yum's cache couldn't expire
yum clean metadata --enablerepo='*'

# install the update candidate
yum install xcp-ng-pv-tools --enablerepo='xcp-ng-updates_testing*'

Here's what you should see in the Windows section of README.txt

We are not allowed to redistribute the windows tools built and signed by
Citrix, so they are not included here.

If you are using Xen Orchestra, you can switch the "Windows Update tools"
advanced parameter on from the "Advanced" tab of the VM view. This will
allow you to install the tools directly from Windows Updates.

Any help in building the tools from the sources to add them back in this
ISO image is welcome. See https://github.com/xcp-ng/xcp/issues/23

olivierlambert created this issue in xcp-ng/xcp

closed Windows PV tools #23

Bungo

Ran, tested, all as expected. Thanks.

ruskofd

I installed this RC1 version via a fresh install using the ISO image (from Proxmox). I used EXT3 thin provisioning as usual (one day Citrix will give us the opportunity to use LVM Thin ).

No problem at all ! Great ! XenOrchestra 5.22 works great, same for XCP.

However, I have not yet had the opportunity to test the USB Passthrough. And finally, my network card does not support SR-IOV, so no opinion on it.

olivierlambert

@Ruskofff LVM thin won't happen I'm afraid. They had to many issues doing it when they tried (data corruption). The path they are using not for thin pro on block based (eg iSCSI) is GFS2. But this require a cluster working, and this part isn't Open Source…

buzzzo

@olivierlambert the fact they didn't succeded doesn't means necessary it does not work.
I'm still surprise they didn't add for ages what you add in few months (ceph, gluster, zfs)

olivierlambert

It's maybe not impossible, but you can't count on Citrix to achieve it, at least with their current priorities.

One of our objective is to master smapiv3 (which is the new storage stack of XS) to be able to build something thin on LVM (amongst other things), but this will require to hire people with advanced knowledge in both XAPI and SR. As you can imagine, there isn't a lot of places where you can find those.