Delaying the unpin of data post-archive, is this possible?

Hello,

Reading through the documentation (particularly https://codimd.web.cern.ch/UrbDTKOLQbOfBzUkYa3kcA#), the unpin of the data from disk storage happens almost immediately post successful tape archive.

While this is expected in usual operation, I’m actually very keen to explore the concept of delaying this based on some metric so I can read the data back for a limited time. I’m not looking to remove the immutability, however reading back a select portion of data that has been written to tape, without a retrieve, is useful for some tools.

Is there some way to delay this particular event?

Alternately, would it be more sensible to use the WFE to with a second path on the EOS instance, and use the WFE with my own functionality to move data to the archival area by my own policy?

Indeed this is how CTA has been designed. The disk buffer is really a buffer, not a cache, and data should be deleted as soon as it has been safely archived to tape.

Yes, you could change the behaviour by defining your own workflow in the WFE, but this is not a supported configuration.

1 Like

Thanks @mdavis, that is what I was thinking, but it’s great to have confirmation.

It does overlap with other desires we have, so I’ll look further into WFE functionality, and as you had mentioned in the past, I’m looking closer at FTS’s bringonline service as a model.