Hello everybody, I had to get out of that multi-tailed multi-threaded big discussion and will try here to summarize it. Now, there was much more to it, but I'll concentrate here on how to handle the new db2 contribution.
I don't intend to neglect the rest, it's just that this is our immediate priority. Unfortunately, we brought the discussion up at a very inopportune moment, just before we're about to do the new release which has already been way too much delayed. But I'm glad that all those concerns were brought up. It's certainly going to benefit the project. The situation we got in is largely due to a strange combination of events. At one point, I could've foreseen what will result, but obviously didn't recognize it. When Holger posted his new db2hadr, Florian asked whether the new feature could be merged into the existing agent (which is only logical). Eventually, I managed to misunderstand what was being folded into what until seeing the final result, which was that the functionality of the existing agent got into the new RA. What is to be done with the new db2 RA? There are four options: a) install under different name (db2-ng or such) b) install under different provider c) replace the old one d) include it after the release There's been strong opposition to both a) and b), though at times some people voted for those options too (I think). No doubt that either of these solutions would be ugly in this case, but hey we needed to do something, right :) I have been against the c) option and IIRC other people have also expressed caution now and then. I am not in a very good position to make an informed decision here: not a db2 expert and the old agent has been obliterated, so I can't even see what has changed. But getting a completely new code immediately before the release is just too dangerous. The new agent is obviously superior to the old one, but the old one has been in use for a while. In fact, there has been a remarkably small number of fixes applied in the last 5 years. I wonder if that means that it was bug free or simply that not many people used it. Also, it was obviously inherited from some heartbeat R1 release, so it does have a long history. So, given that the set of changes is too intrusive and that nobody's entirely happy with the first three options, let's delay the inclusion of the new db2 agent until after the 1.0.4 release. In the meantime, we should try to get an ocft regression test and I just heard that we may be able to get at least one more tester. Agreed? Cheers, Dejan P.S. Thanks to Florian Haas for constructive conversation :) _______________________________________________________ Linux-HA-Dev: Linux-HA-Dev@lists.linux-ha.org http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev Home Page: http://linux-ha.org/
