You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Rucio doesn't ensure that the quota will not be exceeded for an RSE. Even distribution of data for tapes relies on the dm_weight attribute. This attribute is set based on amount of free space, the relative free space and the pledge of tapes. It's a number between 1 and 100 for each RSE. As long as it's non-zero, there's a chance WM will choose this tape as destination based on dm_weight. I'm proposing to set this attribute to 0 if the relative free space in a tape is below 1%.
Use Case
IN2P3 has reached to 99.2% occupancy but since dm_weight was still non-zero, it kept receiving data. We had to switch it off by disabling the wmcore_output_tape attribute.
Possible Solution
If occupancy is 99% for a tape RSE, set dm_weight to 0.
Related Issues
Open questions:
What if all tapes have less than 1% free space?
Should the threshold be in terms of TBs?
Before creating rules, should WM check if there's enough space at the RSE?
Should rucio check if there's enough space at RSE before submitting requests to FTS? [1]
Enhancement Description
Rucio doesn't ensure that the quota will not be exceeded for an RSE. Even distribution of data for tapes relies on the
dm_weight
attribute. This attribute is set based on amount of free space, the relative free space and the pledge of tapes. It's a number between 1 and 100 for each RSE. As long as it's non-zero, there's a chance WM will choose this tape as destination based on dm_weight. I'm proposing to set this attribute to 0 if the relative free space in a tape is below 1%.Use Case
IN2P3 has reached to 99.2% occupancy but since dm_weight was still non-zero, it kept receiving data. We had to switch it off by disabling the
wmcore_output_tape
attribute.Possible Solution
If occupancy is 99% for a tape RSE, set
dm_weight
to 0.Related Issues
Open questions:
@ericvaandering @Panos512 @amaltaro thoughts?
[1] rucio/rucio#7371
The text was updated successfully, but these errors were encountered: