NearNodeFlash / NearNodeFlash.github.io

View this document https://nearnodeflash.github.io/
Apache License 2.0
6 stars 4 forks source link

Investigate capacity obtained vs capacity requested #206

Open bdevcich opened 3 months ago

bdevcich commented 3 months ago

From Bill Loewe:

For XFS rabbit testing, I'm not able to allocate as much capacity as requested. For example, job f4uyWGibPnw requests #DW jobdw type=xfs name=iotest capacity=576GB but only show 483GB: /dev/mapper/31fbd23b--ebd3--4fb8--92d7--5526b9af75c1_0-lv--0 483G 483G 252K 100% /mnt/nnf/15067813-d386-418d-a71d-39b2f2cbe669-0. Is there a means of increasing the capacity to be consistent with the request?

Additionally, for tests that need 224GB per node, we were find with using a capacity of 250GB, but with recent changes this now needs to be a capacity of 300GB to get a filesystem that can support the 224GB.

This deems to be a problem for users where they are trying to smartly allocate only what they need, but the filesystem comes up short of their needs.

bdevcich commented 3 months ago

@matthew-richerson did mention a scaling factor in the NnfStorageProfiles which would explain the 250GB -> 300GB change.