Fun timing, as we've been trying to lift ourselves off of 2.2 and into
2.3. Our Norbert has been hard at work with these efforts.
It's important to me that we do the best we can to stay in line with the
rest of you doing great work closer to the tip.
While I go looking at the changelog on my own, any "notable" things from
the previous 2.x version to go to 2.4 that would jump out? I know we
have the procedure store moving into a region (rolling upgrade, forward
only). The book doesn't have any 2.4 upgrade considerations.
Anything else I (or our operators) should read/care about?
On 3/17/21 4:48 PM, Andrew Purtell wrote:
I would like to propose we update the 'stable' release pointer, currently
pointing at 2.3.4, to 2.4.2.
In my testing with aggressive chaos and ITBLL (but in, unfortunately, due
to resource constraints, in small cluster settings of approximately 10
nodes) 2.4.2 is very stable.
Our sister project Phoenix has updated their build system to support
building against 2.4.1 and later, and the stability of their unit and
integration test suite is not impacted by any known HBase issue.
If there is other criteria that should be considered, I'd like for us to
discuss it. Does there need to be public acknowledgement of a production
user? At scale? (How would we know?) Would you like me to attempt an
at-scale test? On the order of 100 nodes might be possible? If so, what
should be the test scenario and criteria for success? What distinguishes
2.3.x (2.3.4) from 2.4.x (2.4.2) at this point? What would be the area(s)
of concern with respect to moving the stable pointer forward?
