olivierlambert nice catch
It's following a dependencies update, removing an old one ( node-fetch ) . The fix should be merged this morning ( with the ability to resume an import ) https://github.com/vatesfr/xen-orchestra/pull/8440

fbeauchamp opened this pull request in vatesfr/xen-orchestra

closed feat(v2v): resume an incomplete import #8440

so that is probably only a off by one error in the task code
Thanks andrew

Andrew nice catch andrew I will look into it
is it keeping disk attached to dom0 ? (in dashboard -> health )

McHenry xo-cli backupNg.getLogs --json limit='json:500' should work ( the command line parameter are considered as string )

vkeven we don't have ( for now) the feature to create bucket directly from XO. Also I think it is more secure if XO don't know at all the credits of the bucket admin

katapaltes hi, I do think we didnt handle this case.
To be fair, the sheer breadth of the capabilities of vmware is always impressive

Would you be able to show us the VM metadata (especially the .vmx and .vmsd ) to see how we can detect them ? We'll probably won't be able to read this snapshot data for now, so we'll have to at least document a reliable process

On the fast clone side : did you try the checkbox on the bottom (fast clone) ? It should be pretty fast since no data are copied.
On the native snapshot side : I now it's on our roadmap, bu can't give you ETA

vmpr there is a high chance that we improve the scheduler , so maybe be able to plan more precisely when the full occurs, but we will always keep one full backup as the beginning of the chain

You can mitigate the risk on infinite schedule by using health check : restoring the backup automatically after the job, this ensure that , at the time of backup, the files are correct, and that if an issue occurs in the backup you can start a new chain, thus detecting a backup storage issue before needed it after having a production issue

Tristis Oris thanks , I missed a file
I pushed it just now

Tristis Oris that is already a good news.

I pushed an additional fix : the NBD info was not shown on the UI

Tristis Oris no it's on our end

Could you retry nbd + target a block based directory ?
ON my test setup, with the latest changes I get better speed than master ( 190MB/s per disk vs 130-170 depending on the run and settings on master)

I got quite a huge variation between the same runs (40MB/s)

vmpr the oldest one is always a full
a delta is the difference between this backup and the previous, thus we need

You could do infinite delta, without making full, but this comes with a risk : if anything goes wrong ( on backup or storage) , you'll lose the full backup chain

Would it be possible to store a shorter chain here , and use a mirror backup to do longer retention on a cheaper storage ?

Tristis Oris I made a little change, can you update (like the last time ) and retest ?

Tristis Oris Do you have the same performance without NBD ?
Does your storage use blocks ?

Tristis Oris ouch that is quite costly
Can you describe which backup you run ?
Can you check if the duration is different (maybe this is a measurement error and not a slow speed) ?

ih Tristis Oris and Davidj 0 I pushed an update to better handle NBD error, please keep us informed

thanks for your tests . I will recheck NBD backup

vkeven you can request a meeting with our sales teams, but technical discussion are better here if possible

john.c the backup on tape is still on the roadmap but won't be for this step

The main change is that we'll have to rework part of the code to ensure we write data sequentially + the management tools ( which backup is on which tape )

You can try the FS on tape , but the performance penalty is heavy

vkeven as julien-f said, the backup will continue. It will probably be marked as interrupted or successfull with missing VM/ VM info , depending on the phase where the proxy lost connection to the XOA

I think having an autonomous proxy Vs the "dumb" proxy we run today will add a lot of complexity.

Note that you can deploy multiple XOA running backup and a master XOA that can mount all the remote , and then restore any VM . It is not advised to have multiple XOA run the backup on the same VM.

vkeven the architecture didn't change on the proxy side
but since your last message, you may have seen backup health check ( 5.70) , backup immutability ( 5.91 ) encryption at rest (5.74)
object storage is supported since 2021