Marcel Zapf

@derekbit yes a delta rebuild is exactly what I expect with full I meaning a delete and complete copy which is the point of destroying the ssd.

@derekbit This is the problem, expected is delta but longhorn writes everything on the node from scratch

> > [@derekbit](https://github.com/derekbit) This is the problem, expected is delta but longhorn writes everything on the node from scratch > > Hmm, curious, how do you know Longhorn writes everything...

I can also take a look to get smartctl working in prom, then we can get 100% sure what longhorn writes to the ssd by monitoring TBW. I dont know...

And again, after 1.9.0 update   @derekbit The picture is from the third one, because I tried again for testing. Also take a look to the IO monitoring above...

@mantissahz This is with log-level trace after a reboot of the first worker node. For easyer debuging you can focus on the pvc: `pvc-831eaeae-6074-429b-a28e-04b99f932022` which also sits on the worker1....

@mantissahz I've tryed again on worker2 which holds the container with the `pvc-831eaeae-6074-429b-a28e-04b99f932022` this time it uses a full rebuild for the hole disk after a reboot. Only if you...

@mantissahz I'm not sure what the MySQL test deployment is supposed to demonstrate exactly. Once deployed, this deployment will only require around 60 to 100 MB of space and thus...

Hi @mantissahz I can reboot the nodes again and provide you with the supportbundles. The biggest volume is `pvc-831eaeae-6074-429b-a28e-04b99f932022` with around 400GB, so you can look on this volume. This...

@mantissahz any news regarding this issue? Had after some changes in my cluster again over 2TB of writes on the NVMEs because of this longhorn issue :/