Posts Tagged ‘Percona Server’

Speaking at Percona Live Europe Amsterdam

I’m happy to speak at Percona Live Europe Amsterdam 2016 again this year (just look at the awesome schedule). On my agenda:

I’m also signed up for the Community Dinner @ Booking.com, and I reckon you should as well – only 35 spots remain!

Go ahead and register now. You should be able to search Twitter or the Percona blog for discount codes :-)

Learnings from Swift becoming opensource

Swift is now opensource, and it’s interesting to see Craig Federighi talk about it. This is Apple doing right, considering FaceTime is long overdue to being an open standard. People are nitpicking on Apple’s Open Source tagline, but really, this is akin to nitpicking on Mark Zuckerberg donating 99% of his Facebook stock to his new limited liability corporation charity (key: don’t look a gift horse in the mouth).

Apple has chosen to put Swift on Github, and they’ve ensured that it wasn’t just an initial commit, but you’re seeing a lot of history. And it’s the right choice clearly, for engagement — 1,275 watching, 18,884 stars, 2,139 forks, 51 pull requests currently, but most interestingly a lot of accepted code. Even simple things along the lines of “fixing typos” (see commits, eg. d029f7e5ae84cf8f6c12907f9ed0ac0a694881aa, e8b06575d26a684f415af95143ec576a6aa5168d, etc.). 

Swift has open source documentation — like all good open source projects are supposed to have. They use Sphinx and its in the source tree. This is something I’d wish to see from MySQL (docs copyright Oracle, online, but you can take it offline too via PDF) or MariaDB (friendly licensed Knowledge Base), but so far only Percona Server has gotten this right.

What else did Swift do right? Focus on user contributions — the Contributing page is a breath of fresh air. And don’t forget the code of conduct for participating in the project.

But besides just the documentation and contribution pages, I learned something new from one commit in particular — the existence of nproc, part of coreutils. I immediately hopped onto IRC to chat with Nirbhay (our resident MariaDB Galera Cluster expert), because in scripts/wsrep_sst_xtrabackup.sh, we do this via a get_proc() function. We should be focusing on modernising/standardising our codebase, shouldn’t we?

There is a lot we at MariaDB Corporation and the MariaDB Foundation can learn from Swift being opensource and how Apple deals with the community at large. Here’s hoping we get the best practices from it and implement it in due time.

Ubuntu Online Summit: MySQL & Variants in 16.04

I personally have always enjoyed the Ubuntu Developer Summits (UDS), but nowadays they have been converted to the Ubuntu Online Summits (UOS). Attending them is not always convenient (timezone issues, might be travelling, etc.) so I watched the recorded video of a session I was interested in: MySQL & Variants in 16.04.

My key takeaways

  1. Ubuntu 16.04 Xenial Xerus is an LTS release.
  2. The term “cross-grade” is used a lot (it is not about downgrading/upgrading, but being able to use MySQL or MariaDB or Percona Server interchangeably)
  3. It would be nice to see MySQL 5.7 in this release (for Xenial as well as Debian Stretch). From Oracle there is a new packager taking over the task (Lars)
  4. MySQL 5.5 is still the default in Debian, and there needs to be upgrades tested between 5.5 to 5.7 (it looks like the ideal jump is that Ubuntu will not be seeing MySQL 5.6)
  5. Percona Server 5.7 is 60-90 days out; xtrabackup has had some new modifications and deserves an upgrade
  6. Boost is a new requirement for MySQL 5.7 & Percona Server 5.7; some old TokuDB problems in the builds are likely already fixed in MariaDB Server so this can be inherited
  7. MariaDB is waiting to iron out the bugs in 10.0, and may stick to that

My “raw” transcribed notes

  • Attendees:

    • Jon Grimm (Engineering Director for Ubuntu)
    • Robie Basak (Ubuntu)
    • Otto Kekäläinen (MariaDB Foundation)
    • Lars Tangvald, Norvald H. Ryeng (Oracle)
    • George Ormond Lorch III (Percona)
  • Robie: Waiting in Debian for a transition slot from MySQL 5.5 to MySQL 5.6. There’s some discussion with bugs, re: Akonadi, need to also resolve ABI issues with MySQL 5.6. Not really discussed MySQL 5.7 yet.

  • Norvald: 5.7, changes to installation. Client library ABI cleaned up. There may be some clients breaking because of that. No more exported symbols. See: The Client Library, Part 1: The API, the Whole API and Nothing but the API & The Client Library, Part 2: The Version Number
  • mysql_install_db is now replaced by --initialize in the server, so have to rewrite the post-install scripts. Might also have some AppArmour changes. Spoke to people @ DebConf (so best place is to put AppArmour profiles upstream (i.e. in mysql) and Debian and other distros will get it from there). AppArmour profile is in the MySQL source package now. Probably can get away with doing everything as cmake variables.
  • MySQL 5.7 has disabled the old password hashing algorithm, so if people haven’t upgraded they might have problems; so a manual intervention to fix their accounts.
  • Going from MySQL 5.7 to MySQL 5.6? It is done by dump and restore. There is no testing automated downgrades. Are there disk format changes? Norvald is not aware of any. If you use virtual columns in 5.7, you can’t downgrade easily to 5.6.
  • Robie would prefer to not release 5.6 and 5.7 concurrently. During Trusty, there was some level of user confusion. Debian – release team would prefer to see one transfer than two, so is it better to just do a single transition to 5.7?
  • Norvald says there hasn’t been testing from 5.5 -> 5.7. They only support upgrades from 5.5 -> 5.6 -> 5.7. For Ubuntu the choice can be to have 5.6 and then later do 5.7, but Jessie only just released with 5.5, so Stretch with 5.6 might not be a great idea (so users migrating from Jessie to Stretch will go from 5.5 to 5.7). Could also have 5.7 depend on a stripped 5.6 binary (like the embedded server; this is for localhost and the security team shouldn’t be too annoyed) for people to do an upgrade. Norvald says this has not been tried and there needs to be a migration path tested from 5.5 -> 5.7.
  • Conclusion: 5.7 in Stretch. Xenial is an LTS release, and 5.7 should be targeted for that.
  • If the maintainer script fails (postinstall script fails – don’t leave apt in a weird state). If it fails then upgrades, leave a debconf critical notice to say that the service is disabled and then fix it manually. Otto says that leaving /etc in a broken state is terrible, so we should avoid it.
  • Do we (Oracle) have the resources for 5.7 packaging and how soon can it be done in time for Xenial? There were patches from Lars in the git tree, but there haven’t been more recently. Lars will take over the 5.7 transition so if there is a list of work items, this will be settled (Lars will take over from Norvald).
  • There will be a separate session with Norvald/Lars/Robie outside of UOS about 5.7. Defer the Boost conversation after the session as well.
  • George: Percona is mainly looking out towards the 5.7 work and what kind of resources that will be put to that. There are new folk @ Percona to help with this. Percona inherits so much from the upstream codebase, it just works for Percona Server. There is Percona XtraDB Cluster and Percona xtrabackup, and xtrabackup has moved on quite a bit since the last upload (since last November 2014). So might be good idea to look at a refresh. There has also been a lot of work done on Percona XtraDB Cluster and there are some developments with Codership, so they are unsure if they will have their own Percona XtraDB Cluster 5.7 by the time Ubuntu is supposed to ship. When Percona is ready for something, just give Robie a shout to ensure that things happen. 60-90 days before a Percona Server 5.7 release. Just be aware of feature freeze for Xenial.
  • Norvald mentions that Percona Server 5.7 will also depend on Boost and there needs to be a decision on this. George mentions that TokuDB is now part of Percona Server, and it has some of its own requirements as well. Do we include TokuDB? It has requirements like it will only run on 64-bit platforms. Things to figure out going forward? MariaDB has been carrying TokuDB last November, but Robie remembers disabling it in Ubuntu. George says there were some licensing issues back then but they seem to be taken care of.
  • Otto says the builds for TokuDB was failing. It has a dependency on jemalloc, and that might have been the reason there were failures (says George). There may be something else where it doesn’t build on Ubuntu builders. But Otto says that there was a commit where this got fixed about last month. George will follow on, just to absorb it, since the legwork is already complete.
  • Otto: Trusty has 5.5, and Jessie and all other Ubuntu releases have 10.0, and 10.1 was released last month and I’m not quite pushing it to Debian quite yet. Fix 10.0 build fixes, upstream them, then only focus on 10.1. Blocking? (last summer) 5.6 is not in testing, so could not depend on it/changes done in 5.6 mysql-common. Here’s hoping that mysql-common going forward will be generated separately.
  • Robie will take an action to resolve the delta (probably just drop it). To sync MariaDB 10.0 to Xenial.
  • Discussion on /var/lib/mysql/*.flag thing on the list — conclusion at: mailing list — goal: within a single Ubuntu release, people can “cross-grade” between MySQL variants. The goal is to support all 3, and users want to try them, and thats when the bug reports come. Robie’s goal: move to a per-variant data directory. Otto says that once directory names change, 3rd party tools might have breakage. So a working prototype. Migration path is difficult. Maybe the best is to turn /var/lib/mysql into a symlink and store the data elsewhere. PostgreSQL does per version directories today; so studying that is going to happen.

Rackspace Cloud High Availability Databases for MariaDB, MySQL, Percona Server

Continuing on with the cloud theme, I think its worth noting that since mid-2014, Rackspace has offered MariaDB (as well as MySQL and Percona Server) in the cloud, as part of their Cloud Databases offering. It’s powered by OpenStack.

Now there is an additional “High Availability instance” being offered — this gives you up to two replicas per database instance, you have the ability to load balance reads across all replicas (pretty standard), but the cool thing to try out: failover is automatic. It’s not just that if the master fails, you get a new slave being the master; you get a replacement node being added, so as to ensure that your load keeps up with the traffic. These instances don’t cost much more (the higher the memory size, the cheaper it gets — 1.5% extra for something production ready, down to 7.7% more expensive for something to kick around the tires with)

There is also scheduled backups (daily incremental, weekly full) and you can specify the backup window.

Previously on Rackspace, you not only had to spin up a cloud database, but also a compute instance to access your databases. Now, they’re allowing you to get a public IP address, via an ACL.

In another post, I’ll go thru these services with the intention to update my deck and also share the results here. Have you tried or do you use Rackspace Cloud Databases?

Sunsetting HPCloud, whom contributed to making MySQL better

Recently at Percona Live Amsterdam I gave a talk titled Databases in the Hosted Cloud (I’m told I got a 4/5 rating for this talk). It was before AWS re:Invent, so obviously some of the details in the talk have changed. For one, now there is also Amazon RDS for MariaDB. But there has also been other changes, i.e. HP’s Public Cloud (HP Helion Public Cloud) will sunset January 31 2016.

Databases in the Hosted Cloud - Percona Live Amsterdam.001

That’s a slide from my deck. I basically have to caution users as to what’s going on in the cloud world when it comes to their databases. And this one slide shows news reports about HP possibly wanting to exit the cloud world back in April 2015. See: HP Comes to Terms With the Cloud, HP: We’re not leaving the public cloud, and of course the HP blog post from Bill Hilf: HP Helion Strategy to Deliver Hybrid IT Continues Strong.

The tune has of course changed in October 2015: A new model to deliver public cloud. I find this to be quite sad considering they were all very gung ho about pushing OpenStack forward several OSCONs ago. I know many people who made this happen (many ex-MySQL’ers went on to HP to work on OpenStack). I can only feel for them. I guess their important work continues in OpenStack as a whole and all this ends up being part of the HP Helion private cloud.

I think its also worth noting the improvements that Percona Server 5.5 received thanks to HPCloud to make it easier to manage in the cloud:

This pretty much leaves only Rackspace Cloud Databases as being a large OpenStack based offering of databases in the public cloud space, doesn’t it?

HPCloud offered 3 Availability Zones (AZs) per region, and had 2 regions — US-East (Virginia) and US-West. It’s worth remembering that US-West was the only place you could use the Relational DB MySQL service. You also got Percona Server 5.5. You enjoyed 50% off pricing while it was in public beta. 

All this is basically over. Here’s wishing the team well, a big thanks to them for helping make MySQL better and in case you’re looking for more articles to read: H-P Winds Down Cloud-Computing Project

Per query variable settings in MySQL/Percona Server/WebScaleSQL

Recently there was a discussion on the webscalesql mailing list started by Chip Turner on a proposed change to the MAX_STATEMENT_TIME patch. This feature has been known as per query variable settings (WL#681) and even shipping in Percona Server 5.6 as per-query variable statement.

This feature has piqued my interest since 2009, when the MySQL project (then owned by Sun Microsystems) participated in Google Summer of Code 2009, and we got code from Joseph Lukas to do just that (see his tree on Launchpad – lp:~jlukas79/+junk/mysql-server).

So code has been floating around since 2009. It never made it into a shipping release of any MySQL-based distribution till 24 October 2013 when Percona Server 5.6.14-62.0 was released. Percona’s syntax implementation was as suggested in WL#681. This got me curious as to if a feature is already shipping in a distribution of MySQL, what is the WebScaleSQL answer to things – is there a look at other branches or is compatibility from a user/DBA perspective only with focus on upstream?

I got my answer from Steaphan Greene. Very sensible, and a great direction to see how the companies involved can influence upstream MySQL and quite obviously the downstream distributions. It is of course great to note that this syntax improvement will probably be in MySQL 5.7.5 DMR (it already is in 5.7.4 DMR).

For what it’s worth, this feature works well alongside server-side statement timeouts, which Percona Server 5.6 implements (as an alpha quality feature) via the Twitter patch of Davi Arnaut. The MySQL team at Oracle has of course been listening, and in MySQL 5.7.4 DMR (release notes) they too have implemented this feature (WL#6936). Kudos!

Update (6 May 2014): Morgan Tocker has opened up mysql#72540.


i