Converting an OLAP database to TokuDB, part 3: operational stuff

October 14, 2013

This is the third post in a series of posts describing our experience in migrating a large DWH server to TokuDB (see 1st and 2nd parts). This post discusses operations; namely ALTER TABLE operations in TokuDB. We ran into quite a few use cases by this time that we can shed light on.

Quick recap: we've altered one of out DWH slaves to TokuDB, with the goal of migrating most of out servers, including the master, to TokuDB.

Adding an index

Shortly after migrating our server to TokuDB we noticed an unreasonably disproportionate slave lag on our TokuDB slave (red line in chart below) as compared to other slaves.

tokudb-slave-lag

Quick investigation led to the fact that, coincidentally, a manual heavy-duty operation was just taking place, which updated some year's worth of data retroactively. OK, but why so slow on TokuDB? Another quick investigation led to an apples vs. oranges problem: as depicted in part 1, our original setup included MONTHly partitioning on our larger tables, whereas we could not do the same in TokuDB, where we settled for YEARly partitioning.

The heavy-duty operation included a query that was relying on the MONTHly partitioning to do reasonable pruning: a WHERE condition on a date column did the right partition pruning; but where on InnoDB that would filter 1 month's worth of data, on TokuDB it would filter 1 year.

Wasn't it suggested that TokuDB has online table operations? I decided to give it a shot, and add a proper index on our date column (I actually created a compound index, but irrelevant).

It took 13 minutes to add an index on a 1GB TokuDB table (approx. 20GB InnoDB uncompressed equivalent):

  • The ALTER was non blocking: table was unlocked at that duration
  • The client issuing the ALTER was blocked (I thought it would happen completely in the background) -- but who cares?
  • I would say 13 minutes is fast

Not surprisingly adding the index eliminated the problem altogether.

Modifying a PRIMARY KEY

It was suggested by our DBA that there was a long time standing need to modify our PRIMARY KEY. It was impossible to achieve with our InnoDB setup (not enough disk space for the operation, would take weeks to complete if we did have the disk space). Would it be possible to modify our TokuDB tables? On some of our medium-sized tables we issued an ALTER of the form: Continue Reading »

On MySQL plugin configuration

October 1, 2013

MySQL offers plugin API, with which you can add different types of plugins to the server. The API is roughly the same for all plugin types: you implement an init() function, a deinit(); you declare status variables and global variables associated with your plugin, and of course you implement the particular implementation of plugin call.

I wish to discuss the creation and use of global variables for plugins.

Consider the following declaration of a global variable in audit_login:

static MYSQL_SYSVAR_BOOL(enabled, plugin_enabled, PLUGIN_VAR_NOCMDARG,
"enable/disable the plugin's operation, namely writing to file", NULL, NULL, 1);

static struct st_mysql_sys_var * audit_login_sysvars[] = {
    MYSQL_SYSVAR(enabled),
    NULL
};

The above creates a new global variables called "simple_login_audit_enabled": it is composed of the plugin name (known to be "simple_login_audit" in our example) and declared name ("enabled"). It is a boolean, defaults to 1, and is associated with the internal plugin_enabled variable.

Once this variable is declared, you can expect to be able to: Continue Reading »

Introducing audit_login: simple MySQL login logfile based auditing

September 17, 2013

audit_login is a simple MySQL login auditing plugin, logging any login or login attempt to log file in JSON format.

It seems that audit plugins are all the rage lately... We've developed out simple plugin a month ago as part of our database securing efforts; by auditing any login or login attempt we could either intercept or later investigate suspicious logins.

However we quickly realized there is much more to be gathered by this info.

In very short, you install this plugin onto your MySQL server, and your server starts writing into a text file called audit_login.log entries such as follows:

{"ts":"2013-09-11 09:11:47","type":"successful_login","myhost":"gromit03","thread":"74153868","user":"web_user","priv_user":"web_user","host":"web-87.localdomain","ip":"10.0.0.87"}
{"ts":"2013-09-11 09:11:55","type":"failed_login","myhost":"gromit03","thread":"74153869","user":"backup_user","priv_user":"","host":"web-32","ip":"10.0.0.32"}
{"ts":"2013-09-11 09:11:57","type":"failed_login","myhost":"gromit03","thread":"74153870","user":"backup_user","priv_user":"","host":"web-32","ip":"10.0.0.32"}
{"ts":"2013-09-11 09:12:48","type":"successful_login","myhost":"gromit03","thread":"74153871","user":"root","priv_user":"root","host":"localhost","ip":"10.0.0.111"}
{"ts":"2013-09-11 09:13:26","type":"successful_login","myhost":"gromit03","thread":"74153872","user":"web_user","priv_user":"web_user","host":"web-11.localdomain","ip":"10.0.0.11"}
{"ts":"2013-09-11 09:13:44","type":"successful_login","myhost":"gromit03","thread":"74153873","user":"web_user","priv_user":"web_user","host":"web-40.localdomain","ip":"10.0.0.40"}
{"ts":"2013-09-11 09:13:51","type":"successful_login","myhost":"gromit03","thread":"74153874","user":"web_user","priv_user":"web_user","host":"web-03.localdomain","ip":"10.0.0.03"}
{"ts":"2013-09-11 09:14:09","type":"successful_login","myhost":"gromit03","thread":"74153875","user":"web_user","priv_user":"web_user","host":"web-40.localdomain","ip":"10.0.0.40"}
{"ts":"2013-09-11 10:55:25","type":"successful_login","myhost":"gromit03","thread":"74153876","user":"web_user","priv_user":"web_user","host":"web-87.localdomain","ip":"10.0.0.87"}
{"ts":"2013-09-11 10:55:59","type":"successful_login","myhost":"gromit03","thread":"74153877","user":"web_user","priv_user":"web_user","host":"web-12.localdomain","ip":"10.0.0.12"}
{"ts":"2013-09-11 10:55:59","type":"failed_login","myhost":"gromit03","thread":"74153878","user":"(null)","priv_user":"(null)","host":"(null)","ip":"10.0.0.1"}

In the above your MySQL server is on gromit03, and is accepting connections from other hosts; some successful, some not. What kind of information can you gather from the above?

  • You can tell how many connections are being created on your server
  • Where they came from
  • Where 'root' connections come from
  • Port scans (see last row) can be identified by no credentials. These don't have to be port scans per se; any telnet localhost 3006 followed by Ctrl+D will show the same. Typically these would be either load balancer or monitoring tools checks to see that the 3306 port is active.
  • You can tell which accounts connect, and how many times
  • And you can infer which accounts are stale and can be dropped -- if an account does not connect within a week's time, it's probably stale (pick your own timeframe)

The above is quite interesting on one host; but we have dozens. We've installed this plugin on all our MySQL servers, and we use logstash to aggregate them. We aggregate to two destinations: Continue Reading »

Converting an OLAP database to TokuDB, part 2: the process of migration

September 9, 2013

This is a second in a series of posts describing our experience in migrating a large DWH server to TokuDB. This post discusses the process of migration itself.

As a quick recap (read part 1 here), we have a 2TB compressed InnoDB (4TB uncompressed) based DWH server. Space is running low, and we're looking at TokuDB for answers. Early experiments show that TokuDB's compression could make a good impact on disk space usage. I'm still not discussing performance -- keeping this till later post.

Those with weak hearts can skip right to the end, where we finally have a complete conversion. You can also peek at the very end to find out how much 4TB uncompressed InnoDB data is worth in TokuDB. But you might want to read through. The process was not smooth, and not as expected (it's a war story thing). Throughout the migration we got a lot of insight on TokuDB's behaviour, limitations, conveniences, inconveniences and more.

Disclosure: I have no personal interests and no company interests; throughout the process we were in touch with Tokutek engineers, getting free, friendly & professional advice and providing with input of our own. Most of this content has already been presented to Tokutek throughout the process. TokuDB is open source and free to use, though commercial license is also available.

How do you convert 4TB worth of data to TokuDB?

Obviously one table at a time. But we had another restriction: you may recall I took a live slave for the migration process. And we wanted to end the process with a live slave. So the restriction was: keep it replicating!

How easy would that be? Based on our initial tests, I extrapolated over 20 days of conversion from InnoDB to TokuDB. Even with one table at a time, our largest table was expected to convert in some 12-14 days. Can we retain 14 days of binary logs on a server already running low on disk space? If only I knew then what I know today :) Continue Reading »

Three wishes for a new year

September 4, 2013

Another new year by Jewish calendar. What do I wish for the following year?

  1. World peace
  2. Good health to all
  3. Get auto-vacuuming, disk space reclaiming InnoDB tablespaces

No one likes rebuilding huge InnoDB tables. Rebuilds take time, effort, system resources and loss of sleep. I recently rebuilt a 300GB table to realize it reduced to a mere 45GB. How about some background automation?

My wishes in previous two years [2010], [2011], [2012] have not come true. I'm still willing to settle for two out of three.

Converting an OLAP database to TokuDB, part 1

September 3, 2013

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 »

MySQL security top wish list

August 29, 2013

Security seems to have no boundaries. I've been tightening our database security lately, and it seems like this could go on forever: from app to console to privileges to server, there are so many aspects to managing database security. Unfortunately, this is a field where MySQL is in particular weak, and with very little work done in the many years I've been working with MySQL.

My very own top-wanted security features for MySQL follows. Surely this is but a small subset, your mileage may vary.

Autherntication-only SSL

By default, MySQL client API is unencrypted and passwords are sent in cleartext. MySQL supports SSL, but it an "all or nothing" deal: if you want to use SSL, then everything goes by SSL: any query, SELECT, DDL and whatnot.

[UPDATE]: Thanks to Davi & Jan for correcting me on this: passwords are not sent via cleartext. I'm not sure by which constellation I saw cleartext passwords being sent -- but obviously that was long time ago. Just verified via tcpdump, got "mysql_native_password" message and no cleartext password. Lesson learned!

Roles

Need I elaborate? This is a fundamental construct in a database grant system. The effort of maintaining multiple accounts with similar/identical privileges is overwhelming. (PS I haven't used Securich to date)

Host aggregation

In MySQL the combination of user+host makes for a distinct account. Thus, 'gromit'@'192.168.%' is a completely different account than 'gromit'@'10.10.%'. I get the idea: you can have more privileges to, say, gromit@localhost than for gromit@'192.%'. In practice, this only makes a headache. In all my years, I have never encountered nor designed a privilege set where two accounts of the same user had different set of privileges. Never ever ever. It is confusing and pointless: if an account has a different set of roles, just call it by another name! Continue Reading »

Trick: recovering from "no space left on device" issues with MySQL

August 23, 2013

Just read Ronald Bradford's post on an unnecessary 3am (emergency) call. I sympathize! Running out of disk space makes for some weird MySQL behaviour, and in fact whenever I encounter weird behaviour I verify disk space.

But here's a trick I've been using for years to avoid such cases and to be able to recover quickly. It helped me on such events as running out of disk space during ALTER TABLEs or avoiding purging of binary logs when slave is known to be under maintenance.

Ronald suggested it -- just put a dummy file in your @@datadir! I like putting a 1GB dummy file: I typically copy+paste a 1GB binary log file and call it "placeholder.tmp". Then I forget all about it. My disk space should not run out -- if it does it's a cause for emergency. I have monitoring, but sometimes I'm hoping to make an operation on 97%-99% utilization.

If I do run out of disk space: well, MySQL won't let me connect; won't complete an important statement; not sync transaction to disk -- bad situation. Not a problem in our case: we can magically recover 1GB worth of data from the @@datadir, buying us enough time (maybe just minutes) to gracefully complete so necessary operations; connect, KILL, shutdown, abort etc.

common_schema 2.2: better QueryScript isolation & cleanup; TokuDB; table_rotate, split params

August 13, 2013

common_schema 2.2 is released. This is shortly after the 2.1 release; it was only meant as bug fixes release but some interesting things came up, leading to new functionality.

Highlights of the 2.2 release:

  • Better QueryScript isolation & cleanup: isolation improved across replication topology, cleanup done even on error
  • Added TokuDB related views
  • split with "index" hint (Ike, this is for you)
  • table_rotate(): a logrotate-like mechanism for tables
  • better throw()

Drill down:

Better QueryScript isolation & cleanup

common_schema 2.1 introduced persistent tables for QueryScript. This also introduced the problem of isolating concurrent scripts, all reading from and writing to shared tables. In 2.1 isolation was based on session id. However although unique per machine, collisions were possible across replication topology: a script could be issued on master, another on slave (I have such use cases) and both use same (local) session id.

With 2.2 isolation is based on server_id & session id combination; this is unique across a replication topology.

Until 2.1, QueryScript used temporary tables. This meant any error would just break the script, and the tables were left (isolated as they were, and auto-destroyed in time). With persistent tables a script throwing an error meant legacy code piling up. With common_schema 2.2 and on MySQL >= 5.5 all exceptions are caught, cleanup is made, leaving exceptions to be RESIGNALled.

TokuDB views

A couple TokuDB related views help out in converting to TokuDB and in figuring out tables status on disk: Continue Reading »

Tool of the day: q

August 8, 2013

If you work with command line and know your SQL, q is a great tool to use:

q allows you to query your text files or standard input with SQL. You can:

SELECT c1, COUNT(*) FROM /home/shlomi/tmp/my_file.csv GROUP BY c1

And you can:

SELECT all.c2 FROM /tmp/all_engines.txt AS all LEFT JOIN /tmp/innodb_engines.txt AS inno USING (c1, c2) WHERE inno.c3 IS NULL

And you can also combine with your favourite shell commands and tools:

grep "my_term" /tmp/my_file.txt | q "SELECT c4 FROM - JOIN /home/shlomi/static.txt USING (c1)" | xargs touch

Some of q's functionality (and indeed, SQL functionality) can be found in command line tools. You can use grep for pseudo WHERE filtering, or cut for projecting, but you can only get so far with cat my_file.csv | sort | uniq -c | sort -n. SQL is way more powerful for working with tabulated data, and so q makes for a great addition into one's toolbox.

The tool is authored by my colleague Harel Ben-Attia, and is in daily use over at our company (it is in fact installed on all production servers).

It is of course free and open source (get it on GitHub, where you can also find documentation), and very easy to setup. Enjoy!

 
Powered by Wordpress and MySQL. Theme by openark.org