Standardize KV backend responses and handling #496

engelsanchez opened this Issue Feb 26, 2013 · 2 comments


None yet

3 participants


Currently, there seems to be no consensus on what kind of responses KV backends send and how they are handled. We have issue #385 with bad CRC detection blocking writes, for example. Also, we advertise {ok, not_found} as a possible response, but handle {error, not_found} although in that case those should be two very different scenarios (not found because it's not listed or due to an error, say found in keydir but not on file in bitcask), etc. It would be good to have this tricky discussion now and see if we can come up with a standard that any kind of backend can adhere to.

@engelsanchez engelsanchez added a commit that referenced this issue Feb 26, 2013
@engelsanchez engelsanchez Adding bad_crc specifically on read before write
Although plenty of other situations may arise where an error upon
reading should simply allow the write to happen, we're going to wait to
standardize backend responses properly before trying to fix those.
This is just to be prudent until the consequences of doing that are
better understood. This makes the fix specific to bitcask bad CRC
errors and nothing more.
Standardization will hopefully happen in issue #496
jtuple commented Mar 24, 2014

Is this still something we care about / want to do? In any case, assigning to 2.1 cycle.

@jtuple jtuple added this to the 2.1 milestone Mar 24, 2014
jrwest commented Mar 24, 2014

Think its good to put on 2.1 for now, but in that cycle we should either fix it or close as "wontfix"

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment