Rockstor considerations - btrfs "failure" after 3.8-12 upgrade because of Django/python dependencies

Flyer · March 17, 2016, 8:21am

Hi guys, considering I’m still testing Rockstor (for a stable system i won’t use testing updates) cos’ I’d like to run it like the stable nas environment in my office, here are my conclusions/suggestions:

Rockstor seems to be a good system for BTRFS managed via a web ui (and probably the unique one now available over internet), but having btrfs system (snapshots and scheduled tasks) based on a python/django/eggs exposes it to a various number of possible bugs/problems (example: on 3.8-12 now i have a 1 minute/25 snapshots max task, but cos of bugs now i see 380 snapshots, errors telling me that snapshots didn’t create - but they’ve been created -, etc)

Don’t you think it would be better to make btrfs tasks reside on simple bash/sh scripts instead of python (probably this should be less agile, but more stable)

M.

suman · March 17, 2016, 4:25pm

I am sure we can achieve better stability with existing design patterns. But things do break in testing a bit more than we’d like. I am yet to look into this error, but it’s possible that an upstream update to python related packages may have caused it. Thanks for actively testing and reporting these issues, greatly appreciate your contribution.

Flyer · March 18, 2016, 8:42am

@suman @phillxnet

Hope this would help on debug about django/snapshots/etc etc

From web ui you get last snapshots from postgresql db - and you get the same via sql query (see figure)

But from shell ls you can see that snapshots go on

So after the upgrade problem affects to db updates to