Introducing gh-ost: triggerless online schema migrations

I’m thoroughly happy to introduce gh-ost: triggerless, controllable, auditable, testable, trusted online schema change tool released today by GitHub.

gh-ost now powers our production schema migrations. We hit someĀ serious limitations using pt-online-schema-change on our large volume, high trafficĀ tables, to the effect of driving our database to a near grinding halt or even to the extent of causing outages. With gh-ost, we are now able to migrate our busiestĀ tables at any time, peak hours and heavy workloads included, without causing impact to our service.

gh-ost supports testing in production. It goes a long way to build trust, both in integrity and in control.Ā Are your databases just too busy and you cannot run existing online-schema-change tools? Have you suffered outages due to migrations? Are you tired of babysitting migrations that run up to 3:00am? Tired of being the only one tailing logs? Please, take a look at gh-ost. I believe it changes online migration paradigm.

For a more thorough overview, please read the announcement on the GitHub Engineering Blog, and proceed to the documentation.

gh-ost is open sourced under the MIT license.

The mystery of MySQL 5.6 excessive buffer pool flushing

I’m experimenting with upgrading to MySQL 5.6 and am experiencing an unexplained increase in disk I/O utilization. After discussing this with several people I’m publishing in the hope that someone has an enlightenment on this.

We have a few dozens servers in a normal replication topology. On this particular replication topology we’ve already evaluated that STATEMENT based replication is faster than ROW based replication, and so we use SBR. We have two different workloads on our slaves, applied by two different HAProxy groups, on three different data centres. Hardware-wise, servers of two groups use either Virident SSD cards or normal SAS spindle disks.

Our servers are I/O bound. A common query used by both workloads looks up data that does not necessarily have a hotspot, and is very large in volume. DML is low, and we only have a few hundred statements per second executed on master (and propagated through replication).

We have upgraded 6 servers from all datacenters to 5.6, both on SSD and spindle disks, and are experiencing the following phenomena: Continue reading Ā» “The mystery of MySQL 5.6 excessive buffer pool flushing”

TokuDB configuration variables of interest

During our experiments I came upon a few TokuDB variables of interest; if you are using TokuDB you might want to look into these:

  • tokudb_analyze_time

This is a boundary on the number of seconds an ANALYZE TABLE will operate on each index on each partition on a TokuDB table.

That is, if tokudb_analyze_time = 5, and your table has 4 indexes (including PRIMARY) and 7 partitions, then the total runtime is limited to 5*4*7 = 140 seconds.

Default in 7.1.0: 5 seconds

  • tokudb_cache_size

Similar to innodb_buffer_pool_size, this variable sets the amount of memory allocated by TokuDB for caching pages. Like InnoDB the table is clustered within the index, so the cache includes pages for both indexes and data.

Default: 50% of total memory

  • tokudb_directio

Boolean, values are 0/1. Setting tokudb_directio = 1 is like specifying innodb_flush_method = O_DIRECT. Which in turn means the OS should not cache pages requested by TokuDB. Default: 0.

Now here’s the interesting part: we are used to tell InnoDB to get the most memory we can provide (because we want it to cache as much as it can) and to avoid OS caching (because that would mean a page would appear both in the buffer pool and in OS memory, which is a waste). So the following setup is common: Continue reading Ā» “TokuDB configuration variables of interest”

Converting an OLAP database to TokuDB, part 1

This is the first in a series of posts describing my impressions of converting a large OLAP server to TokuDB. There’s a lot to tell, and the experiment is not yet complete, so this is an ongoing blogging. In this post I will describe the case at hand and out initial reasons for looking at TokuDB.

Disclosure: I have no personal interests and no company interests; we did get friendly, useful and free advice from Tokutek engineers. TokuDB is open source and free to use, though commercial license is also available.

The case at hand

We have a large and fast growing DWH MySQL setup. This data warehouse is but one component in a larger data setup, which includes Hadoop, Cassandra and more. For online dashboards and most reports, MySQL is our service. We populate this warehouse mainly via Hive/Hadoop. Thus, we have an hourly load of data from Hive, as well as a larger daily load.

There are some updates on the data, but the majority of writes are just mysqlimports of Hive queries.

Usage of this database is OLAP: no concurrency issues here; we have some should-be-fast-running queries issued by our dashboards, as well as ok-to-run-longer queries issued for reports.

Our initial and most burning trouble is with size. Today we use COMPRESSED InnoDB tables (KEY_BLOCK_SIZE is default, i.e. 8). Our data volume sums right now at about 2TB. I happen to know this translates as 4TB of uncompressed data.

However growth of data is accelerating. A year ago we would capture a dozen GB per month. Today it is a 100GB per month, and by the end of this year it may climb to 150GB per month or more.

Our data is not sharded. We have a simple replication topology of some 6 servers. Machines are quite generous as detailed following. And yet, we will be running out of resources shortly: disk space (total 2.7TB) is now running low and is expected to run out in about six months. One of my first tasks in Outbrain is to find a solution to our DWH growth problem. The solution could be sharding; it could be a commercial DWH product; anything that works. Continue reading Ā» “Converting an OLAP database to TokuDB, part 1”

DELETE, don’t INSERT

Have just read INSERT, Donā€™t DELETE by Aaron Brown, and have some lengthy response, which is why I write this post instead of commenting on said post.

I wish to offer my counter thought and suggest that DELETEs are probably the better choice.

Aaron suggests that, when one wishes to purge rows from some table, a trick can be used: instead of DELETEing unwanted rows, one can INSERT “good” rows into a new table, then switch over with RENAME (but please read referenced post for complete details).

I respectfully disagree on several points discussed.

Lockdown

The fact one needs to block writes during the time of creation of new table is problematic: you need to essentially turn off parts of your application. The posts suggests one could use a slave – but this solution is far from being trivial as well. To switch over, you yet again need to turn off access to DB, even if for a short while. Continue reading Ā» “DELETE, don’t INSERT”

Self throttling MySQL queries

Recap on the problem:

  • A query takes a long time to complete.
  • During this time it makes for a lot of I/O.
  • Query’s I/O overloads the db, making for other queries run slow.

I introduce the notion of self-throttling queries: queries that go to sleep, by themselves, throughout the runtime. The sleep period means the query does not perform I/O at that time, which then means other queries can have their chance to execute.

I present two approaches:

  • The naive approach: for every 1,000 rows, the query sleep for 1 second
  • The factor approach: for every 1,000 rows, the query sleeps for the amount of time it took to iterate those 1,000 rows (effectively doubling the total runtime of the query). Continue reading Ā» “Self throttling MySQL queries”

On generating unique IDs using LAST_INSERT_ID() and other tools

There’s a trick for using LAST_INSERT_ID() to generate sequences in MySQL. Quoting from the Manual:

  1. Create a table to hold the sequence counter and initialize it:
    mysql> CREATE TABLE sequence (id INT NOT NULL);
    mysql> INSERT INTO sequence VALUES (0);
    
  2. Use the table to generate sequence numbers like this:
    mysql> UPDATE sequence SET id=LAST_INSERT_ID(id+1);
    mysql> SELECT LAST_INSERT_ID();
    

This trick calls for trouble.

Contention

A customer was using this trick to generate unique session IDs for his JBoss sessions. These IDs would eventually be written back to the database in the form of log events. Business go well, and one day the customer adds three new JBoss servers (doubling the amount of webapps). All of a sudden, nothing works quite as it used to. All kinds of queries take long seconds to complete; load average becomes very high. Continue reading Ā» “On generating unique IDs using LAST_INSERT_ID() and other tools”

How often should you use OPTIMIZE TABLE? – followup

This post follows up on Baron’s How often should you use OPTIMIZE TABLE?. I had the opportunity of doing some massive purging of data from large tables, and was interested to see the impact of the OPTIMIZE operation on table’s indexes. I worked on some production data I was authorized to provide as example.

The use case

I’ll present a single use case here. The table at hand is a compressed InnoDB table used for logs. I’ve rewritten some column names for privacy:

mysql> show create table logs \G

Create Table: CREATE TABLE `logs` (
 `id` int(11) NOT NULL AUTO_INCREMENT,
 `name` varchar(20) CHARACTER SET ascii COLLATE ascii_bin NOT NULL,
 `ts` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
 `origin` varchar(64) CHARACTER SET ascii COLLATE ascii_bin NOT NULL,
 `message` text NOT NULL,
 `level` tinyint(11) NOT NULL DEFAULT '0',
 `s` char(16) CHARACTER SET ascii COLLATE ascii_bin NOT NULL DEFAULT '',
 PRIMARY KEY (`id`),
 KEY `s` (`s`),
 KEY `name` (`name`,`ts`),
 KEY `origin` (`origin`,`ts`)
) ENGINE=InnoDB AUTO_INCREMENT=186878729 DEFAULT CHARSET=utf8 ROW_FORMAT=COMPRESSED KEY_BLOCK_SIZE=8

The table had log records starting 2010-08-23 and up till 2010-09-02 noon. Table status: Continue reading Ā» “How often should you use OPTIMIZE TABLE? – followup”

Views: better performance with condition pushdown

Justin’s A workaround for the performance problems of TEMPTABLE views post on mysqlperformanceblog.com reminded me of a solution I once saw on a customer’s site.

The customer was using nested views structure, up to depth of some 8-9 views. There were a lot of aggregations along the way, and even the simplest query resulted with a LOT of subqueries, temporary tables, and vast amounts of data, even if only to return with a couple of rows.

While we worked to solve this, a developer showed me his own trick. His trick is now impossible to implement, but there’s a hack around this.

Let’s use the world database to illustrate. Look at the following view definition: Continue reading Ā» “Views: better performance with condition pushdown”

Reducing locks by narrowing primary key

In a period of two weeks, I had two cases with the exact same symptoms.

Database users were experiencing low responsiveness. DBAs were seeing locks occurring on seemingly normal tables. In particular, looking at Innotop, it seemed that INSERTs were causing the locks.

In both cases, tables were InnoDB. In both cases, there was a PRIMARY KEY on the combination of all 5 columns. And in both cases, there was no clear explanation as for why the PRIMARY KEY was chosen as such.

Choosing a proper PRIMARY KEY

Especially with InnoDB, which uses clustered index structure, the PRIMARY KEY is of particular importance. Besides the fact that a bloated PRIMARY KEY bloats the entire clustered index and secondary keys (see: The depth of an index: primer), it is also a source for locks. It’s true that any UNIQUE KEY can serve as a PRIMARY KEY. But not all such keys are good candidates. Continue reading Ā» “Reducing locks by narrowing primary key”