@Forza
Thank you for your answer. If I take the "EPYC-path": Did you ever see, that a 2P-system is slower, than its 1P-pendant?

I am using USB-servers from SEH and Silex. Both of them are working.

Silex can provide higher Performance - SEH is IMHO nicer for license-dongels, etc.

@sumansahabd
Its an LVM-repository. To replicate, you need a snapshot first. A snapshot needs at least 2x the size of the volume on the repository.

Please check, if this is really the case. If you have a VHD which consumes 51% of the SR, you are not able to snapshot/replicate.

@florent
Sorry, but next issue:
Backup size is always shown as 122 KiB although, there are 4GB saved to the remote.

If I save to a storage without "many-files"-seting, 4 GiB is shown

@florent
My problem seems to be, that NBD is not getting activated/tried:

Dec 27 09:35:59 XOA xo-server[6477]: 2022-12-27T08:35:59.951Z xo:backups:DeltaBackupWriter INFO use nbd is NOT activated

...although:

• useNbd = true is set
• Backup-Remote is an SMB-share wir the setting "Store backup as multiple data blocks instead of a whole VHD file. (creates 500-1000 files per backed up TB but allows faster merge) "
• NBD-network is activated on one interface, XO and XCP-ng can work with

Is there any other dependency, that stops NBD from getting used?

Just if somebody else is searching for the same thing, although it is not directly related to XOA:

If you want to get these alerts by mail and you are not using XCP-ng-Center, you can activate them with:

xe pool-param-set uuid=pool-uuid other-config:mail-destination=mail@example.com
xe pool-param-set uuid=pool-uuid other-config:ssmtp-mailhub=mailserverip:port

KPS

That's quite interesting!

It makes sense, that the scheduler "rotates" the load over the cores, but it wonders me, that the "single-threaded-load" of the VM is not shown on the whole system.

Shouldn't one core be at least at 75%, if one core in the VM is at 75%?

@olivierlambert
Thank you for your answer. I am really looking forward to get the "task-system".

Currently, it's quite hard to see the status of the system. For me (and at least most of your competitors), the merge-process is part of a backup-job, as it produces I/O and takes bandwidth...

Hi!

When I start a backup in XOA, I can see a part of the progree in the "tasks"-section. But: This seems to be quite limited.
Is there any possibility to see the "real" progress?

One example:
A delta-job is shown as "Status:Successful" before the merge did complete.
--> Where can I see the status of the merge-process

Thank you for your help
KPS

@fred974
It would be great, if you could write down some lines about the issue and how it could get fixed

@olivierlambert

I can confirm: This issue is solved in XO Lite, but still present in XOA.

@olivierlambert
As it seems to be harder, than I did hope it would be: Would it be an option to open an external VNC-application:

• xenstore-read /local/domain/XXX/console/vnc-port
  --> This could give you IP and port for the VNC connection

@olivierlambert
I found an old bug report from 2016: https://github.com/vatesfr/xen-orchestra/issues/1426

...and some threads in the forum:
https://xcp-ng.org/forum/topic/2385/xo-console-keyboard-on-ios-and-ipados-no-input-possible
https://xcp-ng.org/forum/topic/2004/console-window-as-input-field

Sorry, but I did not see these posts prior to open that new thread.

But: It would be really nice, if there would be the option to define console as "input field".

Best wishes!
KPS

fufroma created this issue in vatesfr/xen-orchestra

open Virtual keyboard for console #1426

Hi!

There are some situations, where I need to access XOA "to go" on my mobile phone.
While it is great to be able to e.g. start and stop VMs, I sometimes need to use the VM-console.

On iOS, the keyboard is "faded out", when the console is opened. I am only able to use the "mouse".
Are you aware of any possiblity to use the keyboard in the XOA-VM-console?

Thank you and best wishes
KPS

Hi!

For every VM, the number of sockets/cores can be configured, but did you ever see any impact through the NUMA change?

Let's take an AMD Epyc CPU, that provides 4 NUMA nodes.
I want to tune the performance of a Windows/Linux-VM, that has 8 vCPUs assigned.

Options:
--> 1 socket, 8 cores
--> 2 sockets, 4 cores each
--> 4 sockets, 2 cores each
--> 8 sockets, 1 cores each

What would you prefer, or doesn't this have any performance impact?

Thank you for your thoughts
KPS

Hi!

Advanced Live Telemetry looks great, but takes the nodes of an XCP-ng-cluster separated.

When I migrate a VM from one host to another, data is not consistent anymore.

• The "souce-host" does not show historic data for the VM anymore.

• ListenpunktThe "destination-host" can only show data after the migration.

Is there any possibility to merge the data clusterwise?

Best wishes
KPS

Hi!

Is anybody here already using the new AMD Epyc Genoa generation?
According to Citrix, they should be fully supported.

Are you happy with their performance?

Best wishes
KPS