feat: built-in periodic gc #87
Conversation
hacdias
force-pushed
the
gc
branch
2 times, most recently
from
2d434bb to
05330a8
Compare
There was a problem hiding this comment.
Generally LGTM, found a small bug and it seems like we can probably abstract over some of the OS layer specifics.
Adding tests here seems annoying, but wonder if we should have a basic one. For example set the GC interval to 1s and free 100% and make sure the data gets GC'd from the blockstore layer (nothing we can really do about the DB layer if that GC happens separately).
|
@aschmahmann I refactored with the package you mentioned for cross-os compatibility. I also added a test, but it is far from what I would want. We need a way to actually start an isolated rainbow instance with all configurations. That's tracked in #89. |
the server handler registers global prometheus stuff. if we do that twice, panics. we have to figure out how to do this without globally setting prometheus
|
@aschmahmann can you also take a look at ipfs/boxo#579? There was a race condition in the tests that that PR fixes. |
Co-authored-by: Adin Schmahmann <adin.schmahmann@gmail.com>
Closes #65.
DEPENDS ON ipfs/boxo#579
Matches (I think) the behaviour of our current script: https://github.com/ipshipyard/waterworks-infra/blob/master/ansible/roles/rainbow/files/rainbow-gc.sh
Every hour, calls GC if less than 30% of the disk space is available.
Only works on Unix based systems.