Moreover, I have also tested rollback from 2.4 to 1.6 and after taking care of some rsgroup ordering issues as part of the downgrade, we have seen smooth downgrade as well (with the only exception that last RS that stays on 2.4 needs to be shutdown non-gracefully or killed, because of HBASE-17931).
On Mon, 11 Oct 2021 at 1:10 PM, Viraj Jasani <vjas...@apache.org> wrote: > I have done some chaos testing as well as ingestion testing 3-4 times on > the recent 2.4 versions, each spanning the duration of around 8-10 hr. And > also tested rolling upgrade from 1.6 to 2.4 a couple of times after > applying the recent fixes on adding missing CFs in meta table as part of > the upgrade. (Upgrade requires upgrading all RS first before masters) > > Things seem to be running well so far, no significant concerns have been > identified with default configs so far. > > Andrew and I will keep testing this release line on different > infrastructure with different tooling, but for now, I have sufficient test > data to provide my +1 for moving the stable pointers to 2.4. And thank you > for starting this thread, Sean! > > > On Mon, 11 Oct 2021 at 1:46 AM, Sean Busbey <bus...@apache.org> wrote: > >> Hi folks! >> >> I'd like us to move the stable pointer from 2.3.z releases to 2.4.6+ >> releases. >> >> The last time we talked about doing so[1] there was a desire to get >> some documentation together on what we as a community expect from the >> "stable" release line. We have an issue tracking those needs[2], but >> AFAICT we haven't had sufficient community interest to get criteria >> together over the last 6 months. >> >> The 2.3 release line started in July 2020 and there is active >> discussion about declaring it EOL[3]. 2.4 releases have been going >> since Dec 2020 and there is attestation that 2.4 has been as or more >> stable than 2.3 in a testing environment[4]. >> >> Personally, I view the "stable" pointer as simply a way to say "if you >> are new to our community we'd like you to use this release." I >> personally think 2.4.z currently meets that standard and we need not >> block updating on something more rigorous. >> >> What do folks think? >> >> -busbey >> >> [1]: >> "[DISCUSS] Updating the 'stable' pointer to 2.4.2" : >> https://s.apache.org/6cz3t >> >> [2]: >> https://issues.apache.org/jira/browse/HBASE-25690 >> >> [3]: >> "[DISCUSS] EOL 2.3" : https://s.apache.org/pgkge >> >> [4]: >> Message on thread "[DISCUSS] EOL 2.3" >> https://s.apache.org/ks7wk >> >
