Forum OpenACS Development: Re: OpenACS 5.8.0 + PostgreSQL 9.3

Collapse
4: Re: OpenACS 5.8.0 + PostgreSQL 9.3 (response to 1)
Posted by Jim Lynch on 09/20/13 03:59 AM
Ok, maybe you did it already.... Seeing pretty often the incredible amount of effort you put into releases, I've come to expect some occasional heroism on your part. I'm pretty sure that you and your team are responsible for keeping openacs alive. In any case, let's find out:

When you first released 5.8, what version of postgres did you have running? Was it that version of pg which you first run the core tests on?

Did you run those core tests again with 5.8 on any other postgres, and if so which, and were there any differences between those results, and if so do you find them significant? Would you be willing to summarize at least one pair of results you compared?

Collapse
5: Re: OpenACS 5.8.0 + PostgreSQL 9.3 (response to 4)
Posted by Gustaf Neumann on 09/20/13 08:49 PM
not sure, what you mean by "heroism" and "keeping alive", but according to our download statistics [1], OpenACS 5.8.0 is doing very well (in 3 weeks more than 7.500 downloads, this are more than the total number of downloads of OpenACS 5.4.0, 5.4.1, 5.4.2, 5.4.3, 5.5.0, 5.5.1). OpenAcs is doing quite well. Since we use OpenACS in large production environments, we have certainly interest on a good code quality.

Concerning bugs etc.: The most important contribution of OpenACS 5.8.0 are that it runs on pg 9.2+ without any necessary configuration changes, and it runs efficiently also on large sites (using e.g. recursive queries for permissions). Before that, large sites had to use pg 8.3, which is already at EOL. Without the change, OpenACS would have been doomed on the longer range.

Concerning pg-versions: we run the regression tests (automated testing on acs-core) on the following pg releases: 8.4.16, 9.0.10, 9.1.5, 9.2.3, 9.2.4, all with the same results (8 errors in spell-check, complains about non-configured spellcheck, an entry about this is active in the bug tracker) and two complaints from acs-authentication (auth_password_recover, auth_password_retrieve; tries to send mail, which is on the test machines not configured). As i wrote earlier, these are exactly the same errors we were seeing with OpenACS 5.6.0 and 5.7.0 ... and now with pg 9.3.0.

The regression test does certainly not cover all of OpenACS. Here are two more figures concerning bug fixing that one might find interesting:

- this year, so far 78 bugs from the bug-tracker were fixed/resolved/closed, this is more than the sum from the previous 4 years (between 5 and 20 per year)

- OpenACS.org runs the freshest code from the oacs-5-8 branch and uses not only acs-core, but as well several packages, in which we fixed many bugs not contained in the bugtracker. These changes led to much lower numbers of error entries in the error.log: 100 days ago (using aolserver + OpenACS 5.5 + pg 8.3.5) we saw abut 17.000 errors per day, last week, we were at 3000, yesterday at 428 (many of the remaining errors are from from invalid urls, etc.). Several bugs were site-related and a single bug could cause many errors/day, but we are feeding all changes back to the oacs-5-8 branch. Eventually this branch will comes out as a OpenACS 5.8.1 release (don't ask me when).

hope, this helps
-gn

[1] https://openacs.org/projects/openacs/download/

Collapse
6: Re: OpenACS 5.8.0 + PostgreSQL 9.3 (response to 5)
Posted by Jim Lynch on 09/24/13 02:34 PM
That all sounds great... in the last few days, I went from being convinced openacs was dying or dead... now, I see it's very much alive; thanks for that.

And yes, you've completely answered my question about whether pg version is influencial over test results; according to your message, it's not influencial as the results are the same for 8.4 and all the 9.x pgs running when you've run the regression tests. Thanks for that too.

-Jim