As a bystander (mostly), I've followed the related discussion in past.
Please ignore me if this doesn't contribute to anything, since I'm
obviously missing a lot of context that has resulted in some of the
discussion seem less fruitful than expected. Anyway, just looking at the
matter at hand, I'd really like to see improvements implemented in the
code base. I'm quite often seeing weird stuff happening with Zookeeper,
and while everything usually gets sorted after a timeout or a restart or
so, the mechanism could certainly be more robust.
As I see it, what you've done, Mark, is pretty extensive. And I
personally find large change sets difficult to handle. Would it be
possible to split all the changes into more manageable pieces so that
they could be reviewed and merged gradually? It'd be sad to just let all
the work contribute to nothing.
Best,
Ere
Mark Miller kirjoitti 1.9.2022 klo 4.15:
There was too much diversion to switch to it in the end. Just a sample of
changes: I rewrote the entire Overseer and collections API implementation,
rewrote ZkStateReader, rewrote all the Zookeeper management with a
recursive snide watcher strategy, made all the primary paths async with
async IO, moved lots of parts to offheap bytebuffer management, rewrote
most of JavaBin, changed almost all collection classes to fast util,
rewrote all the thread management, reworked most of the distributed update
code…
Even if that was not the start of a list but the full list, it’s well
beyond the scope of what others could consume even if you take as a given
that all of my choices are universal agreed with. Factor in that those
using Solr today are generally making it do what they need it to do (even a
typical major upgrade is general not an interesting proposition) and things
have come to a stop before the hards parts have even began.
Kind of a let down I know, but selfishly, I just really wanted to see it.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]