trident icon indicating copy to clipboard operation
trident copied to clipboard

Configurable volume to accommodate LUN metadata

Open P0lskay opened this issue 11 months ago • 4 comments

Describe the solution you'd like Currently, Trident defaults to allocating 10% of PV volume to LUN metadata. Some users deploy Trident PVs with sizes greater than 100GB In this case, it seems that allocating 10% of the volume for LUN metadata is unnecessary. At the same time, our team would like to allocate 20% of the LUN for metadata, as we often work with smaller PVs and want to ensure data protection. To address all these issues at once, we suggest allowing users to configure the amount of space allocated for LUN metadata within their volumes.

P0lskay avatar Jan 15 '25 04:01 P0lskay

Just to better understand the use case: What drives your requirement to have 20% for LUN metadata? From my experience it should never be that much?

wonderland avatar Jan 20 '25 08:01 wonderland

Just to better understand the use case: What drives your requirement to have 20% for LUN metadata? From my experience it should never be that much?

We want to allocate 20% of the available space on the LUN for metadata because we have experienced situations where the discard mount option was not enabled at the operating system level. When working with small disk sizes (up to 100 GB), the LUN was approximately 99% full, and when a large load was applied to the disk, the LUN would overflow, causing the volume to overflow and input to stop. Allocating 20% for LUN metadata would protect us from these scenarios.

P0lskay avatar Jan 21 '25 09:01 P0lskay

I have a similar problem where it seems that LUN metadata of 10% is not enough to handle a filesystem being filled (https://github.com/NetApp/trident/issues/903). Having the filesystem being full should not trigger error on the storage array side.

cfillot avatar Mar 06 '25 07:03 cfillot

I have a similar problem where it seems that LUN metadata of 10% is not enough to handle a filesystem being filled (#903). Having the filesystem being full should not trigger error on the storage array side.

@wonderland As you can see, this is not an individual case.

P0lskay avatar Mar 06 '25 10:03 P0lskay