XCP-ng 8.3 betas and RCs feedback 🚀

ThierryC01

@florent Sure, here are the screenshots after importing the config.json saved from 8.2 before the upgrade. The backup Jobs page was empty.

Xcp-ng backup_1.png

Xcp-ng backup_2.png

And this is the files on my remote share

Xcp-ng backups.png

florent

@ThierryC01 is the remote connected ? what happen when you click on "refresh backup list" ?

edit : it seems you have multiple remotes, each one with only one vm ?

ThierryC01

@florent Remote is connected all the time, it is a NAS and each VM has its own folder otherwise it is just a mess of mumble jumble folder names that no one knows which VM it is. Refreshing the backup jobs or restore page does nothing.

ThierryC01

@stormi I edited the fstab to mount my ISO SR and it is back online. Next time, I will remember to backup this file too before any upgrade.

florent

@ThierryC01 said in XCP-ng 8.3 beta :

@florent Remote is connected all the time, it is a NAS and each VM has its own folder otherwise it is just a mess of mumble jumble folder names that no one knows which VM it is. Refreshing the backup jobs or restore page does nothing.

ok, that's a first for me
The only way this can work is with multiples remote and job set in XO. Are they still there ? no error ?

ThierryC01

@florent I found the source of the problem, I already had it when I was trying the backup feature in 8.2 and it is with the remote settings: if there is a logon issue, there is no way to edit the password, save and go with it, you have to re-create the remote entirely which is extremely frustrating. For some reason, the password encryption does not behave as expected, when, editing the password, it is replaced by "obfuscated-xxxxxxx unique ID xxxxxxx" so it cannot work.

The message I get when trying to connect to a remote folder or just after editing the password and save
Xcp-ng remote_1.png

The obfuscated password displayed when editing
Xcp-ng remote_3png.png

And the error log for the logon attempt with the correct password:

remote.test
{
  "id": "365964ae-ea3c-41a5-80eb-0c5eb595f01d"
}
{
  "messageName": "session_setup_step2",
  "params": {},
  "message": "STATUS_LOGON_FAILURE (0xC000006D) : The attempted logon is invalid. This is either due to a bad username or authentication information.",
  "code": "STATUS_LOGON_FAILURE",
  "name": "Error",
  "stack": "Error: STATUS_LOGON_FAILURE (0xC000006D) : The attempted logon is invalid. This is either due to a bad username or authentication information.
    at SMB2Forge.request (/etc/xen-orchestra/node_modules/@marsaud/smb2/lib/tools/smb2-forge.js:22:15)
    at /etc/xen-orchestra/node_modules/@marsaud/smb2/lib/tools/smb2-connection.js:108:11
    at Object.<anonymous> (/etc/xen-orchestra/node_modules/@marsaud/smb2/lib/tools/message.js:15:15)
    at Socket.<anonymous> (/etc/xen-orchestra/node_modules/@marsaud/smb2/lib/tools/smb2-forge.js:71:31)
    at Socket.emit (events.js:400:28)
    at Socket.emit (domain.js:475:12)
    at Socket.patchedEmit [as emit] (/etc/xen-orchestra/@xen-orchestra/log/configure.js:118:17)
    at addChunk (internal/streams/readable.js:293:12)
    at readableAddChunk (internal/streams/readable.js:267:9)
    at Socket.Readable.push (internal/streams/readable.js:206:10)
    at TCP.onStreamRead (internal/stream_base_commons.js:188:23)
    at TCP.callbackTrampoline (internal/async_hooks.js:130:17)"
}

florent

@ThierryC01 the obfuscated- part is added to the data sent to the web client and the logs

But , in theory, editing a remote with a new password will change the password.

You can check the data saved in the XO database by executing yarn xo-server-db ls remote from your xo folder

olivierlambert

@ThierryC01 said in XCP-ng 8.3 beta :

mess of mumble jumble folder names that no one knows which VM it is

That's really not a good idea to do one BR per VM, as you discovered. The current folder structure is here for a reason, it's not a "mess of mumble jumble folder names": it's a stable structure meant to be parsed by XO, not by a human.

You are currently fighting/bending against the tool, this has consequences.

ThierryC01

@olivierlambert That is not true, a remote folder is still a remote folder and we should be able to backup any VM wherever we need/want. The issue here is when the login fails once, there is no way to edit the logon password and succeed. It was working very well until I upgraded from 8.2 to 8.3 and then had to restore a VM from backup. No grudge here, just a fact.

What I call a mumble jumble naming of folder is just that every VM is identified by a uuid in the same folder if all backups are stored in the same place which makes it difficult to sort should we need to either use a backup on another platform or delete to regain disk space.

I am just fighting to get a basic feature working: edit the password of a connection without having to delete the record and start over.

ThierryC01

@florent That is the theory yes, but in reality, editing the password fails.

ThierryC01

Okay... I have to apologize for the pollution of this topic, I was running a container version of XOA (Xen Orchestra, commit 422a2) which I use in emergency cases such as my actual up to date XOA VM which crashed due to an LVM issue and was trying to restore. I managed to get my actual XOA VM running and everything is fine, no problem whatsoever, all backups are here using the XOA VM, available to restore and all. I really feel bad for wasting everyone's time. I will update my emergency XOA container from now on, I thought that all information were stored on the Xcp-ng server, good to know.

laurentm

@ThierryC01

About your LVM crash, was it with NVMe U2 SSD ?
Running XenServer from more than ten years and XCP-NG from 7.5, the worst issue I encounted has been recently with a very brand new DELL R650 fitted with some excellent Kingston U2 NVMe DC1500M,
connected straight to the motherboard, without any hardware raid controller. It appears that NVMe can disappear randomly due to some stupid power save feature in Linux kernel. I have patched that and I'm using each ssd separetely, in order to reduce the risk, and without LVM at XCP-NG level, only in my Linux VMs.

olivierlambert

You mean APST?

https://docs.xcp-ng.org/troubleshooting/storage/disappearing-NVMe-drives/

ThierryC01

@laurentm No, for some stupid reason, the LVM volume was 100% full and I could not write any disk configuration nor expand the volume until I got rid of some logs.gz. All this happened during an update of XOA, I think there should be some safeguard in Linux to avoid such situation.

I will install a new XOA VM somewhere else as a backup since the container does not update to the latest commit. I hope they could sync the backup configuration.

florent

@ThierryC01 said in XCP-ng 8.3 beta :

@florent That is the theory yes, but in reality, editing the password fails.

Does it work with a newly created remote ? i.e. you create a new samba, it connects, you edit the password, it still works ?

bufanda

@ThierryC01 What Container Image do you use. I personally use this container

https://github.com/Ezka77/xen-orchestra-ce

It builds weekly and follows the master of xen-orchestra. So latest version is always most updated one.

stormi

@bufanda Let's try to stay on the topic of XCP-ng 8.3 beta

bufanda

@stormi Yeah sure, just thought it may help with the issue that was reported if the container used is outdated.

But anyways to the topic of XPC-NG 8.3beta. Are there any updates planned in the near future it has been a while since I recieved Updates on my test cluster. And while I don't run an AMD Processor on my test cluster there were security fixes for 8.2 but they didn't made to my 8.3 or did the repo config change and I missed that?

stormi

@bufanda So, yes, we have a big update in the pipe, but we still have bits to tweak before we can push it to all testers.

Regarding security updates, we don't offer any guarantee regarding the update delay on the beta, until the final release.

bufanda

@stormi Thx for the info. And I understand that there is no guarantee, was just curious if I had a problem on my cluster or if there was really nothing inbetween. It's a test cluster in the end so no mission critical things running on it anyways.