So the big news for Spetember was the CPAN Testers Server upgrade. I posted progress on the mailing list, as well as via a Questhub quest, and was hoping no-one would be caught unawares, sadly some were. Despite a few initial hiccups the migration from old server to new server went very well. A few interesting lessons were learned, particularly with MySQL and a reminder of how bad SimpleDB is ... again. On the old server I had been using Metabase v0.16, but last year Ricardo released a newer version, with some changes to the API. Not realising this the initial install for downloading metabase reports failed. Thankfully the changes were minor, and fit better with DBIx::Class style of queries. Unfortunately the result was still a problem for SimpleDB. I have reverted back to v0.16 for the time-being, but will be upgrading to the latest version again soon. A big thank you to Chris from Bytemark Hosting, again for helping getting us upgraded.

The biggest difference, now we're on the new server, has been the speed at which pages are built and loaded. With the website pages sitting on a separate SSD disk from the db and all other files, the response times have been significantly improved. This has also meant that the builder is running almost constantly now, and is keeping up very well. The report generator is also running much better, and is getting blocked by disk I/O considerably less, such that we are now roughly 1-2 hours behind the report submissions to the Metabase. Okay so its not realtime (a much wielded stick waved at CPAN Testers), but its not far off it.

Another benefit to the speed improvements, have been to remove all the restrictions on spiders and crawlers. For the last week this has been monitored carefully, and thankfully no-one has been abusing it. One thing that has been interesting to see, is that Google's RSS Feed-Fetcher tells you how many subscribers it has for a particular RSS feed. In most cases it's just 1, typically the distribution author I suspect, but it would be great if other RSS aggregators did the same, and there are several by the looks of things, as it would be an interesting stat to have on the respective pages.

The Admin site has been put on hold while I worked on the server upgrade, but I'll be back on it very soon. Most of the site is complete (see the Questhub quest for details), so it shouldn't be too long now before its live. There are some changes happening in the database, so we can finally link the tester relationships (email addresses, PAUSE and names) between all the reports and websites. This will finally mean that the Tester Leaderboard on the Statistics site can be built with very little duplication. It will also mean that testers can manage their own appearance on the leaderboard too, saving me having to clean up entries manually every few months, or on request from testers.

Ben Bullock created a quest on Questhub to add text search of reports to CPAN Testers. I originally posted that this was already planned. However, after further consideration, I think it would be worth having fresh eyes looking at building a search site. However, the new reports are not text based. We present them as text based on the Reports website, and in emails, but that is not how they are recorded in the Metabase. Reports are now structured data, and any search site should take advantage of that. As such, I've asked Ben if he can look at writing a search site that can incorporate specific searches that can take advantage of this structured data. Ben's original target of December 2014 should definitely be possible, although I suspect Ben will be able to have some results long before this. It would also be an ideal project for next year's QA Hackathon.

Although happening in October, I did want to thank Alceu Rodrigues de Freitas Junior, for adding the page SmokeTesting on the Wiki. If you have any notes that would help others, whether that involves setting up a smoke environment, or hints and tips for authors or testers, it is very much appreciated. The Wiki is open to all to update, so please feel free to add or update entries to keep the site current.

On the mailing list, Ron Savage posted a query about Tests fail but strings identical in a test report. The problem is specific to Win32, and highlights the need to understand how Win32 I/O layers. Tony Cook posted a response explaining how to solve this.

When I firsted announced the server upgrade, Alex Balhatchet asked who pays for the servers. For the CPAN Testers Server, this is currently Birmingham Perl Mongers, but this will likely only be for another 2 years, after which we'll have to start looking for either a dedicated sponsor or dip into the CPAN Testers Fund. It's probably a good opportunity to remind people about the Fund, as although we do have a lot of support for CPAN Testers, we would welcome more to help us stay vibrant for many years to come. If you or your company want to show your appreciation of CPAN Testers, monetary donations are always very welcome. If you think you can help in other ways, please get in touch.

That's all for now, but expect plenty of news next month too.