Xaprb

I’ll be presenting at the Southern Computer Measurement Group’s meeting on Thursday. I’ll discuss how to extract scalability and performance metrics from TCP/IP packet headers. Registration is inexpensive, but it’s even less if you register by Monday. There is a full schedule of other good talks — it is an all-day meeting.

Blackhole tables are often used on a so-called “relay slave” where some operation needs to happen but no data needs to exist. This used to have a bug that prevented AUTO_INCREMENT columns from propagating the right values through replication, but that was fixed. It turns out there’s another bug, though, that has the same effect. This one is caused when there is an INSERT into a Blackhole table, where the source data is SELECT-ed from another Blackhole table.

I think it’s wise to keep it simple. MySQL has tons of cool little features that theoretically suit edge-case uses and make ninja tricks possible, but I really trust the core plain-Jane functionality so much more than these edge-case features. That’s precisely because they often have some edge-case bugs, especially with replication.

Something that’s new to MySQL recently is Galera replication. The more I think about it, the more I think it’s fundamentally the right way to replicate. Statement-based replication was brittle; row-based is less so, but still has all kinds of gotchas. The real problem with both is that they are built into the server, not the storage engine. Engine-level replication is the way to go. PBXT has had engine-level replication for a while, although I’ve never used PBXT in production (and kudos to PostgreSQL for adding built-in replication, too). I used to want InnoDB to do replication via streaming the redo logs and applying them, but that actually has a lot of limitations. Galera is InnoDB’s answer to engine-level replication. I think Galera holds a lot of promise for the future.

For me, next week’s Percona Live is a lot like other events: I know I will be busy meeting and greeting and helping unpack boxes and so forth, but I’ll also be talking and attending other talks. With five concurrent tracks, it’s tough to decide. This is roughly the equivalent of the MySQL conference every year, which has more tracks, but historically it’s been easy to cross off some talks as sales pitches. Not so with this event! Here’s what I want to see the most:

• Clustrix’s keynote. I’ve watched the progress of our tests with Clustrix very closely. If you’re coming to the conference, don’t oversleep.
• Henrik’s session on high availability technologies.
• Johan’s session on NDB performance.
• Fred’s talk on Spider.
• Shlomi’s talk about SQL coding. Shlomi’s talks kind of remind me of the things Beat Vontobel does (solving a sudoku with a single query, anyone?)
• And of course, Domas’s closing keynote.

Perceptive readers will note that I left some slots empty. I haven’t decided yet. It’s too hard to choose.

I will be presenting two things next week: a 3-hour tutorial on how to diagnose problems (part one: method; part two: tools), and a session on black-box performance analysis and scalability modeling with TCP packet headers.

See you there — I hope!