Monday, August 15, 2016

General Overview of Multimaster Replication

General Overview of Multimaster Replication 
 
Multimaster replication is a utility that allows data in multiple databases to be automatically kept in sync. For example, in a multimaster replication system, if a row gets inserted into one of the databases in the system, that row will be automatically propagated to all of the other databases in that system. Updates and deletes to the data in any of the databases will be propagated in the same way.
A multimaster replication environment is set up by configuring databases to be part of a “replication group”. One of the databases in the group is defined as the “master definition site,” and all of the other databases in the group are classified as “master sites.” The main difference between the two types of sites is that most of the replication administration commands must be invoked from the master definition site.

There are two basic ways that transactions get propagated to remote databases—“synchronously” and “asynchronously”. Synchronous replication occurs by causing each transaction to be applied to all the master sites in a group immediately. The way this is achieved is by using Oracle’s two- phase commit functionality, to ensure that all of the databases in question can apply a given transaction. If any of the sites in the group cannot accept the transaction (such as because the site’s database has crashed, or the network connection to a database is down) then none of the master sites in the replication group will be able to accept the transaction—the transaction will not be able to take place.
The way asynchronous replication works is that all the transactions that occur on a site are temporarily placed in a buffer, called the “deferred transaction queue,” or deftran queue. Periodically, such as once per minute, all of the transactions in a site’s deftran queue get sent to all of the other sites, by “push” jobs. These jobs get created by calling the “schedule_push” procedure. Finally, the transactions in a deftran queue that have already been sent to other sites must be periodically purged, to prevent the deftran queue from growing too large.

The vast majority of customer sites that use multimaster replication use asynchronous replication rather than synchronous. One of the reasons for this is that asynchronous replication has been available for a much longer time; the initial versions of multimaster replication only allowed for asynchronous propagation. The main reason that asynchronous is used, though, is because it has many advantages over synchronous.

First of all, asynchronous replication uses much less network bandwidth and provides higher performance than synchronous replication. The primary reason for this is that it is more efficient to store multiple transactions and then propagate them all as a group, rather than to propagate each transaction separately.

This is particularly important when the sites in question are very far apart geographically (such as having one site in San Francisco and another in New York). Another reason for these bandwidth and performance improvements is that there is much more overhead associated with synchronous replication because each and every transaction requires that separate connections be established to all of the other sites in the replication group. With asynchronous replication, fewer connections need to be established, since transactions are propagated as a group.

The biggest advantage of asynchronous replication, though, is that it provides for high availability of the replication group. With asynchronous replication, if one of the sites in the replication group crashes, all of the other sites will still be able to accept updates—the transactions that are made on the remaining sites will just “stack up” in those sites’ deftran queues until the down site becomes available.

On the other hand, with synchronous replication, if any one of the sites becomes unavailable (such as because of a database crash or a network failure) then none of the sites will be updatable. This is because with synchronous replication, each and every transaction must be able to be immediately applied to all of the sites in the replication group, and of course if a site is unreachable no transactions will be able to be applied to it. This means that not only does synchronous replication not provide any higher database availability, it can actually provide lower availability than using a single database!

Multi-Master Replication

If You *Must* Deploy Multi-Master Replication, Read This First

An increasing number of organizations run applications that depend on MySQL multi-master replication between remote sites.   I have worked on several such implementations recently.  This article summarizes the lessons from those experiences that seem most useful when deploying multi-master on existing as well as new applications.

Let's start by defining terms.  Multi-master replication means that applications update the same tables on different masters, and the changes replicate automatically between those masters.  Remote sites mean that the masters are separated by a wide area network (WAN), which implies high average network latency of 100ms or more.  WAN network latency is also characterized by a long tail, ranging from seconds due to congestion to hours or even days if a ship runs over the wrong undersea cable.

With the definitions in mind we can proceed to the lessons.  The list is not exhaustive but includes a few insights that may not be obvious if you are new to multi-master topologies.  Also, I have omitted issues like monitoring replication, using InnoDB to make slaves crash-safe, or provisioning new nodes.  If you use master/slave replication, you are likely familiar with these topics already.

1. Use the Right Replication Technology and Configure It Properly

The best overall tool for MySQL multi-master replication between sites is Tungsten.  The main reason for this assertion is that Tungsten uses a flexible, asynchronous, point-to-point, master/slave replication model that handles a wide variety of topologies such as star replication or all-to-all.  Even so, you have to configure Tungsten properly.  The following topology is currently my favorite:
  • All-to-all topology.  Each master replicates directly to every other master.  This handles prolonged network outages or replication failures well, because one or more masters can drop out without breaking  replication between the remaining masters or requiring reconfiguration.  When the broken master(s) return, replication just resumes on all sides.  All-to-all does not work well if you have a large number of masters.  
  • Updates are not logged on slaves.  This keeps master binlogs simple, which is helpful for debugging, and eliminates the possibility of loops.  It also requires some extra configuration if the masters have their own slaves, as would be the case in a Tungsten Enterprise cluster
There are many ways to set up multi-master replication replication, and the right choice varies according to the number of masters, whether you have local clustering, or other considerations.  Giuseppe Maxia has described many topologies, for example here, and the Tungsten Cookbook has even more details.

One approach you should approach with special caution is MySQL circular replication.  In topologies of three or more nodes, circular replication results in broken systems if one of the masters fails.  Also, you should be wary of any kind of synchronous multi-master replication across sites that are separated by more than 50 kilometers (i.e. 1-2ms latency).  Synchronous replication makes a siren-like promise of consistency but the price you pay is slow performance under normal conditions and broken replication when WAN links go down.

2. Use Row-Based Replication to Avoid Data Drift

Replication depends on deterministic updates--a transaction that changes 10 rows on the original master should change exactly the same rows when it executes against a replica.  Unfortunately many SQL statements that are deterministic in master/slave replication are non-deterministic in multi-master topologies.  Consider the following example, which gives a 10% raise to employees in department #35.

   UPDATE emp SET salary = salary * 1.1 WHERE dep_id = 35;

If all masters add employees, then the number of employees who actually get the raise will vary depending on whether such additions have replicated to all masters.  Your servers will very likely become inconsistent with statement replication.  The fix is to enable row-based replication using binlog-format=row in my.cnf.  Row replication transfers the exact row updates from each master to the others and eliminates ambiguity.

3. Prevent Key Collisions on INSERTs

For applications that use auto-increment keys, MySQL offers a useful trick to ensure that such keys do not  collide between masters using the auto-increment-increment and auto-increment-offset parameters in my.cnf.  The following example ensures that auto-increment keys start at 1 and increment by 4 to give values like 1, 5, 9, etc. on this server.

server-id=1
auto-increment-offset = 1
auto-increment-increment = 4
This works so long as your applications use auto-increment keys faithfully.  However, any table that either does not have a primary key or where the key is not an auto-increment field is suspect.  You need to hunt them down and ensure the application generates a proper key that does not collide across masters, for example using UUIDs or by putting the server ID into the key.   Here is a query on the MySQL information schema to help locate tables that do not have an auto-increment primary key. 
SELECT t.table_schema, t.table_name 
  FROM information_schema.tables t 
    WHERE NOT EXISTS 
      (SELECT * FROM information_schema.columns c
       WHERE t.table_schema = c.table_schema  
         AND t.table_name = c.table_name
         AND c.column_key = 'PRI'
         AND c.extra = 'auto_increment')

4. Beware of Semantic Conflicts in Applications

Neither Tungsten nor MySQL native replication can resolve conflicts, though we are starting to design this capability for Tungsten.  You need to avoid them in your applications.  Here are a few tips as you go about this.
First, avoid obvious conflicts.  These include inserting data with the same keys on different masters (described above), updating rows in two places at once, or deleting rows that are updated elsewhere.  Any of these can cause errors that will break replication or cause your masters to become out of sync.  The good news is that many of these problems are not hard to detect and eliminate using properly formatted transactions.  The bad news is that these are the easy conflicts.  There are others that are much harder to address.  
For example, accounting systems need to generate unbroken sequences of numbers for invoices.  A common approach is to use a table that holds the next invoice number and increment it in the same transaction that creates a new invoice.  Another accounting example is reports that need to read the value of accounts consistently, for example at monthly close.  Neither example works off-the-shelf in a multi-master system with asynchronous replication, as they both require some form of synchronization to ensure global consistency across masters.  These and other such cases may force substantial application changes.  Some applications simply do not work with multi-master topologies for this reason. 

5. Remove Triggers or Make Them Harmless
Triggers are a bane of replication.  They conflict with row replication if they run by accident on the slave.  They can also create strange conflicts due to weird behavior/bugs (like this) or other problems like needing definer accounts present.  MySQL native replication turns triggers off on slaves when using row replication, which is a very nice feature that prevents a lot of problems.  
Tungsten on the other hand cannot suppress slave-side triggers.  You must instead alter each trigger to add an IF statement that prevents the trigger from running on the slave.  The technique is described in the Tungsten Cookbook.  It is actually quite flexible and has some advantages for cleaning up data because you can also suppress trigger execution on the master.  
You should regard all triggers with suspicion when moving to multi-master.  If you cannot eliminate triggers, at least find them, look at them carefully to ensure they do not generate conflicts, and test them very thoroughly before deployment.  Here's a query to help you hunt them down: 
SELECT trigger_schema, trigger_name 
  FROM information_schema.triggers;
6. Have a Plan for Sorting Out Mixed Up Data

Master/slave replication has its discontents, but at least sorting out messed up replicas is simple: re-provision from another slave or the master.  No so with multi-master topologies--you can easily get into a situation where all masters have transactions you need to preserve and the only way to sort things out is to track down differences and update masters directly.   Here are some thoughts on how to do this.
  1. Ensure you have tools to detect inconsistencies.  Tungsten has built-in consistency checking with the 'trepctl check' command.  You can also use the Percona Toolkit pt-table-checksum to find differences.  Be forewarned that neither of these works especially well on large tables and may give false results if more than one master is active when you run them.  
  2. Consider relaxing foreign key constraints.  I love foreign keys because they keep data in sync.  However, they can also create problems for fixing messed up data, because the constraints may break replication or make it difficult to go table-by-table when synchronizing across masters.  There is an argument for being a little more relaxed in multi-master settings. 
  3. Switch masters off if possible.  Fixing problems is a lot easier if you can quiesce applications on all but one master.  
  4. Know how to fix data.  Being handy with SQL is very helpful for fixing up problems.  I find SELECT INTO OUTFILE and LOAD DATA INFILE quite handy for moving changes between masters.  Don't forget SET SESSION LOG_FILE_BIN=0 to keep changes from being logged and breaking replication elsewhere.  There are also various synchronization tools like pt-table-sync, but I do not know enough about them to make recommendations.  
At this point it's probably worth mentioning commercial support.  Unless you are a replication guru, it is very comforting to have somebody to call when you are dealing with messed up masters.  Even better, expert advice early on can help you avoid problems in the first place.

(Disclaimer:  My company sells support for Tungsten so I'm not unbiased.  That said, commercial outfits really earn their keep on problems like this.)

7. Test Everything

Cutting corners on testing for multi-master can really hurt.  This article has described a lot of things to look for, so put together a test plan and check for them.  Here are a few tips on procedure:
  1. Set up a realistic pre-prod test with production data snapshots.  
  2. Have a way to reset your test environment quickly from a single master, so you can get back to a consistent state to restart testing. 
  3. Run tests on all masters, not just one.  You never know if things are properly configured everywhere until you try. 
  4. Check data consistency after tests.  Quiesce your applications and run a consistency check to compare tables across masters. 
It is tempting to take shortcuts or slack off, so you'll need to find ways to improve your motivation.  If it helps, picture yourself explaining to the people you work for why your DBMS servers have conflicting data with broken replication, and the problem is getting worse because you cannot take applications offline to fix things.  It is a lot easier to ask for more time to test.  An even better approach is to hire great QA people and give them time to do the job right.

Summary

Before moving to a multi-master replication topology you should ask yourself whether the trouble is justified.  You can get many of the benefits of multi-master with system-of-record architectures with a lot less heartburn.  That said, an increasing number of applications do require full multi-master across multiple sites.  If you operate one of them, I hope this article is helpful in getting you deployed or improving what you already have.

Tungsten does a pretty good job of multi-master replication already, but I am optimistic we can make it much better.  There is a wealth of obvious features around conflict resolution, data repair, and up-front detection of problems that will make life better for Tungsten users and reduce our support load.  Plus I believe we can make it easier for developers to write applications that run on multi-master DBMS topologies.  You will see more about how we do this in future articles on this blog.

Sunday, August 7, 2016

MyRocks vs InnoDB

  • InnoDB writes between 8X and 14X more data to SSD per transaction than RocksDB
  • RocksDB sustains about 1.5X more QPS
  • Compressed/uncompressed InnoDB uses 2X/3X more SSD space than RocksDB
I encourage others to use long running benchmark tests and present IO efficiency metrics in addition to performance results.


Configuration


I used the same configuration as described in the previous post with one difference. For this test I ran 168 iterations of the query step and each step ran for 1 hour. The test ran for 7 days while the previous test ran for 1 day. What I describe as QPS below is TPS (transactions/second) and when I use per query below I mean per transaction. The IO efficiency metrics are measured by iostat. I report the database size in GB at the end of each day - hours 1, 24, 48, 72, 96, 120, 144 and 168. For each one hour interval I collect:
  • average QPS
  • iostat reads per query (r/q)
  • iostat KB read per query(rKB/q)
  • iostat KB written per query (wKB/q)
  • iostat reads per second (r/s)
  • iostat KB read per second (rKB/s)
  • iostat KB written per second (wKB/s)
Tests were run for several binaries:
  • myrocks.zlib - Facebook MySQL 5.6, RocksDB with zlib compression
  • innodb56.none - upstream MySQL 5.6.26, InnoDB without compression
  • innodb57.none - upstream MySQL 5.7.10, InnoDB without compression
  • innodb56.zlib - upstream MySQL 5.6.26, InnoDB with zlib compression
  • innodb57.zlib - upstream MySQL 5.7.10, InnoDB with zlib compression

Better Compression


Compressed InnoDB uses about 2X more SSD space than MyRocks. Uncompressed InnoDB uses about 3.1X more SSD space than MyRocks. This graph shows the database size every 24 hours. Note that the database gets more data as a function of the QPS rate and MyRocks has more data than InnoDB after 168 hours -- myrocks.zlib has 6.4% more rows than inno57.none and 7.2% more rows than inno57.zlib after 7 days.

Better Performance


I'd be happy if MyRocks matched the QPS from InnoDB and only beat it on IO efficiency. But it wins on QPS and IO efficiency. The data below is QPS over time. MyRocks gets at least 1.5X more QPS than compressed InnoDB. It also does a lot better than uncompressed InnoDB but who wants to use 3X more SSD space. The QPS growth at test start for InnoDB with zlib happens because there are stalls until the compressed b-tree pages fragment. I think this is a problem with mutexes in the pessimistic insert/update code for compressed InnoDB.
The graph below shows the average QPS from each 1-hour interval.

Better IO Efficiency


I present IO efficiency metrics here using data from iostat normalized by the QPS rate to show the amount of IO done per transaction.
This result is remarkable. InnoDB writes between 8X and 14X more to storage per transaction than MyRocks. This means that workloads can use TLC SSD with MyRocks when InnoDB requires MLC and that workloads can use SSD with MyRocks when SSD doesn't have sufficient endurance for InnoDB. Running a busy database on SSD is so much easier than using disk.
This result doesn't include the additional write-amplification from flash GC that occurs with InnoDB compared to MyRocks because the MyRocks write pattern generates less work for flash GC. I previously described how that increased the flash write-rate by about 1.5X for InnoDB compared to RocksDB for the device I use. This means that the real difference in write rates on SSD might mean that InnoDB writes 12X to 21X more to storage than MyRocks.

Using iostat metrics from hour 168 InnoDB writes 8.7, 10.5, 8.9 and 10.6 times more to storage per transaction compared to RocksDB. Using iostat data from hour 167 the difference is 11.5, 13.9, 11.7 and 14.0 times more data written.

The graph below shows the number of KB written to SSD per transaction from each 1-hour interval.
This graph shows the number of SSD reads per transaction. MyRocks has the smallest rate. While an LSM can have an IO penalty for range queries and range queries are the most frequent operation in Linkbench, that isn't a problem for this workload.This graph shows the number of KB read from SSD per transaction. The rate for MyRocks is in between the rates for compressed and uncompressed InnoDB. The r/q rate is more important than this rate as long as the difference here isn't extreme. The rate for MyRocks includes reads done for user queries and reads done in the background for compaction.

Absolute iostat results


These graphs show absolute rates from iostat. The data is the average rate per 1-hour interval.

The first graph is the rate for iostat r/s. The rate is larger for MyRocks because it sustains the most QPS.
The next graph shows the rate for iostat wKB/s. Note that InnoDB sustains between 100 and 200 MB/s of writes. The SSD device is much busier doing writes for InnoDB than for MyRocks. IO capacity used for writes isn't available for reads even when endurance isn't an issue. More writes means more erases from flash GC and erases are a source of IO stalls when a read gets stuck behind the erase on the same channel.The last graph shows the rate for iostat rKB/s. The rates are larger for uncompressed InnoDB and MyRocks compared to uncompressed InnoDB. The SSD is very busy if you combine the iostat rates for rKB/s and wKB/s. Votes:

Monday, August 11, 2014

TokuDB : Hot Index Creation

Hot Index Creation

TokuDB allows you to add indexes to an existing table and still perform inserts and queries on that table while the index is being created.

The ONLINE keyword is not used. Instead, the value of the tokudb_create_index_online client session variable is examined. More information is available in TokuDB Variables.
Hot index creation is invoked using the CREATE INDEX command after setting
tokudb_create_index_online=on.

Here's an example:

SET tokudb_create_index_online=ON;
Query OK, 0 rows affected (0.00 sec)

CREATE INDEX index table (field_name);
 
Alternatively, using the ALTER TABLE command for creating an index will create the index offline (with the table unavailable for inserts or queries), regardless of the value of tokudb_create_index_online. The only way to hot create an index is to use the CREATE INDEX command.

Hot creating an index will be slower than creating the index offline, and progress depends how busy the mysqld server is with other tasks. Progress of the index creation can be seen by using the SHOW PROCESSLIST command (in another client). Once the index creation completes, the new index will be used in future query plans.

If more than one hot CREATE INDEX is issued for a particular table, the indexes will be created serially. An index creation that is waiting for another to complete will be shown as Locked in SHOW PROCESSLIST. We recommend that each CREATE INDEX be allowed to complete before the next one is started.

TokuDB : Hot Column Addition and Deletion

Hot Column Addition and Deletion

From 18 hours to 3 seconds!

Hot Column Addition and Deletion (HCAD) Overview

TokuDB v5.0 introduces several features that are new to the MySQL world. In this series of posts, we’re going to present some information on these features: what’s the feature, how does it work under the hood, and how do you get the most out of this feature in your MySQL setup.

Today we start with HCAD: Hot Column Addition and Deletion. Many users have had the experience of loading a bunch of data into a table and associated indexes, only to find that adding some columns or removing them would be useful.

alter table X add column Y int default 0;
 
or the like takes a long time — hours or more — during which time the table is write locked, meaning no insertions/deletions/updates and no queries on the new column until the alter table is done.
Mark Callaghan points out that changing the row format in InnoDB is a “significant project”, so it looked like slow alter tables were going to be a challenge for MySQL for the foreseeable future. Slow alter tables is a reason for the inability of MySQL to scale to large tables.

TokuDB v5.0 changes all that with the introduction of HCAD. You can add or delete columns from an existing table with minimal downtime — just the time for MySQL itself to close and reopen the table. The total downtime is seconds to minutes.

Here we present an example of HCAD in action. See this page for details of the experiment. Drum roll…
TokuDB:

mysql> alter table ontime add column totalTime int default 0;
Query OK, 0 rows affected (3.33 sec)
 
InnoDB:
mysql> alter table ontime add column totalTime int default 0;
Query OK, 122225386 rows affected (17 hours 44 min 40.85 sec)
 
That’s 19,000x faster! Goodbye long downtimes.

As a note, the “0 rows affected” for TokuDB means that the column addition work happens in the background. All queries on the table, however, will see the new column as soon as the alter table returns, in this case after 3.33 sec.

Friday, April 18, 2014

How can I optimize a mysqldump of a large database?

How can I optimize a mysqldump of a large database ?

I have a symfony application with an InnoDB database that is ~2GB with 57 tables. The majority of the size of the database resides in a single table (~1.2GB). I am currently using mysqldump to backup the database nightly.
Due to my comcast connection, oftentimes if I am running a dump manually my connection to the server will timeout before the dump is complete causing me to have to rerun the dump. [I currently run a cron that does the dump nightly, this is just for dumps that I run manually.]
Is there a way to speed up the dumps for the connection timeout issue, but also to limit the time the server is occupied with this process?
BTW, I am currently working on reducing the size of the overall database to resolve this issue.










The main bottleneck in the dump like this is drive I/O. You are reading a load of data and writing it again. You can speed this up in a number of ways:
  • Make sure your output is going to a different drive(s) than the one(s) the database files are stored on - this will make a massive difference with spinning disks as the drive heads will not be constantly flicking between the location being read from and the location being written to.
  • The output of mysqldump will be very compressible, so if you can not separate the output from the input as mentioned above pipe the output through gzip or similar. This will reduce the amount of writing being done (so reduce the overall IO load, and the amount of head movement) at the expense of some CPU time (which you may have a lot of spare at these times anyway). Also, pass the output through a pipe utility (like pv) that supports large write buffers to group blocks written to the drives together more, again to reduce the effect of head-movement latency - this will make quite a difference if using the --quick option to reduce the RAM impact of backing up large tables).
  • Only run your backup process when IO load is otherwise low.
You may be fixing the wrong issue though: it might be easier to address the connection drops instead (though reducing the I/O load imposed by your backups will help reduce the effect you have on other users so is worth trying anyway). Could you run your manual backups through screen (or similar tools like tmux)? That way if your connection to the server drops you can just reconnect and reattach to the screen session without any processes getting interrupted.
If you are sending the data directly over the connection (i.e. you are running mysqldump on your local machine against a remote database, so the dump appears locally) you might be better off running the dump on the server first, compressing as needed, then transferring the data over the network using a tool (such as rsync) which supports partial transfers so you can resume the transfer (instead of restarting) if a connection drop interrupts it.
As part of your "reducing the size of the overall database to resolve this issue" I would guess that a large chunk of your data does not change. You might be able to move a large chunk of the 1.2Gb from that main table off into another and remove that from those that are copied by the mysqldump call. You don't need to backup this data every time if it never changes. Splitting data between tables and databases this way is usually referred to as data partitioning and can also allow you to spread the data and I/O load over multiple drives. High-end database have built in support for automatic partitioning, though in mysql you will probably have to do it manually and alter your data access layer to account for it.
Straying off-topic for this site (so you should probably nip over to ServerFault or SuperUser to ask if you need more detail): If you seem to be losing connections due to inactivity, check the options in your SSH server and SSH client to make sure keep-alive packets are enabled and being sent often enough. If seeing drops even if the connection is active you could also try using OpenVPN or similar to wrap the connection - it should handle a short drop, even a complete drop if your entire connection is down for a few seconds, such that the SSH client and server don't notice.







INSIGHT INTO DOING BACKUPS WITH mysqldump
IMHO Doing backups has become more of an art form if you know just how to approach it
You have options

Option 1 : mysqldump an entire mysql instance
This is the easiest one, the no-brainer !!!
mysqldump -h... -u... -p... --hex-blob --routines --triggers --all-databases | gzip > MySQLData.sql.gz
Everything written in one file: table structures, indexes, triggers, stored procedures, users, encrypted passwords. Other mysqldump options can also export different styles of INSERT commands, log file and position coordinates from binary logs, database creation options, partial data (--where option), and so forth.

Option 2 : mysqldump separate databases into separate data files
Start by creating a list of databases (2 techniques to do this)
Technique 1
mysql -h... -u... -p... -A --skip-column-names -e"SELECT schema_name FROM information_schema.schemata WHERE schema_name NOT IN ('information_schema','mysql')" > ListOfDatabases.txt
Technique 2
mysql -h... -u... -p... -A --skip-column-names -e"SELECT DISTINCT table_schema FROM information_schema.tables WHERE table_schema NOT IN ('information_schema','mysql')" > ListOfDatabases.txt
Technique 1 is the fastest way. Technique 2 is the surest and safest. Technique 2 is better because, sometimes, users create folders for general purposes in /var/lib/mysql (datadir) which are not database related. The information_schema would register the folder as a database in the information_schema.schemata table. Technique 2 would bypass folders that do not contain mysql data.
Once you compile the list of databases, you can proceed to loop through the list and mysqldump them, even in parallel if so desired.
for DB in `cat ListOfDatabases.txt`
do
    mysqldump -h... -u... -p... --hex-blob --routines --triggers ${DB} | gzip > ${DB}.sql.gz &
done
wait
If there are too many databases to launch at one time, parallel dump them 10 at a time:
COMMIT_COUNT=0
COMMIT_LIMIT=10
for DB in `cat ListOfDatabases.txt`
do
    mysqldump -h... -u... -p... --hex-blob --routines --triggers ${DB} | gzip > ${DB}.sql.gz &
    (( COMMIT_COUNT++ ))
    if [ ${COMMIT_COUNT} -eq ${COMMIT_LIMIT} ]
    then
        COMMIT_COUNT=0
        wait
    fi
done
if [ ${COMMIT_COUNT} -gt 0 ]
then
    wait
fi
Option 3 : mysqldump separate tables into separate data files
Start by creating a list of tables
mysql -h... -u... -p... -A --skip-column-names -e"SELECT CONCAT(table_schema,'.',table_name) FROM information_schema.tables WHERE table_schema NOT IN ('information_schema','mysql')" > ListOfTables.txt
Then dump all tables in groups of 10
COMMIT_COUNT=0
COMMIT_LIMIT=10
for DBTB in `cat ListOfTables.txt`
do
    DB=`echo ${DBTB} | sed 's/\./ /g' | awk '{print $1}'`
    TB=`echo ${DBTB} | sed 's/\./ /g' | awk '{print $2}'`
    mysqldump -h... -u... -p... --hex-blob --triggers ${DB} ${TB} | gzip > ${DB}_${TB}.sql.gz &
    (( COMMIT_COUNT++ ))
    if [ ${COMMIT_COUNT} -eq ${COMMIT_LIMIT} ]
    then
        COMMIT_COUNT=0
        wait
    fi
done
if [ ${COMMIT_COUNT} -gt 0 ]
then
    wait
fi
 
Option 4 : USE YOUR IMAGINATION
Try variations of the aforementioned Options plus techniques for clean snapshots
Examples
  1. Order the list of tables by the size of each tables ascending or descending.
  2. Using separate process, run "FLUSH TABLES WITH READ LOCK; SELECT SLEEP(86400)" before launching mysqldumps. Kill this process after mysqldumps are complete. This is helpful if a database contains both InnoDB and MyISAM
  3. Save the mysqldumps in dated folders and rotate out old backup folders.
  4. Load whole instance mysqldumps into standalone servers.
CAVEAT
Only Option 1 brings everything. The drawback is that mysqldumps created this way can only be reloaded into the same majot release version of mysql that the mysqldump was generated. In other words, a mysqldump from a MySQL 5.0 database cannot be loaded in 5.1 or 5.5. The reason ? The mysql schema is total different among major releases.
Options 2 and 3 do not include saving usernames and passwords.
Here is the generic way to dump the SQL Grants for users that is readble and more portable
mysql -h... -u... -p... --skip-column-names -A -e"SELECT CONCAT('SHOW GRANTS FOR ''',user,'''@''',host,''';') FROM mysql.user WHERE user<>''" | mysql -h... -u... -p... --skip-column-names -A | sed 's/$/;/g' > MySQLGrants.sql
Option 3 does not save the stored procedures, so you can do the following
mysqldump -h... -u... -p... --no-data --no-create-info --routines > MySQLStoredProcedures.sql &
Another point that should be noted is concerning InnoDB. If your have a large InnoDB buffer pool, it makes sense to flush it as best you can before performing any backups. Otherwise, MySQL spends the time flushing tables with leftover dirty page out of the buffer pool. Here is what I suggest:
About 1 hour before performing the backup run this SQL command
SET GLOBAL innodb_max_dirty_pages_pct = 0;
In MySQL 5.5 default innodb_max_dirty_pages_pct is 75. In MySQL 5.1 and back, default innodb_max_dirty_pages_pct is 90. By setting innodb_max_dirty_pages_pct to 0, this will hasten the flushing of dirty pages to disk. This will prevent or at least lessen the impact of cleaning up any incomplete two-phase commits of InnoDB data prior to performing any mysqldump against any InnoDB tables.
FINAL WORD ON mysqldump
Most people shy away from mysqldump in favor of other tools and those tools are indeed good.
Such tools include
  1. MAATKIT (parallel dump/restore scripts, from Percona [Deprecated but great])
  2. XtraBackup (TopNotch Snapshot Backup from Percona)
  3. CDP R1Soft (MySQL Module Option that takes point-in-time snapshots)
  4. MySQL Enterprise Backup (formerly InnoDB Hot Backups [commercial])
If you have the spirit of a true MySQL DBA, you can embrace mysqldump and have the complete mastery over it that can be attained. May all your backups be a reflection of your skills as a MySQL DBA.

Monday, April 14, 2014

Backups and Recover

Backups and Recovery

This is the most important task of an database administrator, you must protect your data at all costs, this means regular backups and regular restores even to another system just to check the integrity of those backups. There is no point in putting yourself in a position where you are holding your breathe when a restore is happening only to find out that the backup is corrupt, try if possible to perform regular restores if not then at least you should be performing a disaster recovery test once per year. Not being able to restore could be a disaster for your company and your job.
To check your backups you can use one or more of the below which I have used in the past
  • use a reporting database if the customers don't need real time data and you have the money and time, Production data could be restored every day to this system which is a very good test
  • use a performance test server with Production data, ideal to test releases of your software against Production data which is generally has more volume then a test system, restore perhaps once a week
  • at least perform a DR once per year to prove the backup solution is working, for example you may have forgotten to backup something not only regarding the database but from the systems as well
Backups and restoring
First lets start with a few terms associated with backups
logical backup this type of backup is created by saving information that represents the logical database structures using SQL statements like create database, create table and insert. This type of backup is ideal when you want to upgrade from one version of MySQL to another however it is a slower method of backing up.
physical backup this type of backup is a backup of the actual database files or disk partitions, this type of backup can be very fast to backup and restore.
full backup a full backup is a standalone backup containing everything in the database, this could then be restored on another server. A full backup can be either logical or physical.
incremental backup this type of backup only contains the data that has changed from the last backup. The advantage of this type of backup is that it is faster as there is not some much data to backup, however the disadvantage is that it takes longer to recover.
consistent backup this is a backup at an exact moment in time, generally you shutdown the database (or quiescent mode) then take the backup.
hot backup this type of backup is taken when the database is running, during the backup both reads and writes are not blocked
warm backup this type of backup is taken when the database is running, however reads are not blocked but writes are prohibited from making any modifications to the database.
cold backup similar to a consistent backup as the database is shutdown before the backup begins
point-in-time restore is a restoration of a database to a specified date and time , some databases use a full backup and recovery logs to restore to that point-in-time, others can only use the last full backup which means that data might have to be re-keyed into the system.
As well as obtaining a backup in your maintenance window you should also be aware on how long a restore will take thus to make sure that you meet you SLA agreements during a DR or if you have to recovery a database due to corruption or user error.
The $64,000 question is how often you should take your backups, and this i am afraid depends, so company are happy for once a month backups other may take two backups per day. The answer generally has to come from the business on what they are prepared to lose, amount of data lost or what has to be re-keyed into the system again. If you have a small company that say has to re-key in 20-50 invoices then that's no big deal, however if you have a trading company that many have to re-key in 10's of thousands of entries/trades then that becomes a problem. You have to add the time it takes to restore the system plus the time it takes to recover the system so that users are able to use it, it is this time that you give to the business to make there decision on what is a acceptable time period that the business can be down for, the shorter the time the more money that will have to be thrown at the solution, if you are talking about zero downtime then we would have to implement a high availability solution which could cost a lot of money, if you are happy with 1 days downtime then this should be enough to restore and recovery a database and to re-key in some entries to make the database consistent with the companies paper work.
As you saw above there are a number of ways to backup a database, depending on the the available time to perform a backup will make you decide on what method to use, if you have a short maintenance window with a large database then a incremental backup maybe the only option, but you have a large maintenance window with a small database then you could perform a full backup, remember what ever option you use with have a impact on the recovery time.
One point to make is that you backups should be taken off-site if held on tape or copied across to an other system in another location, if an incident happened on the original system for example a fire you don't want to lose your backups as well, the storing of off-site data should be part of you DR plan.
Enough of talking about backups lets see how you can actually take one, there are a number of backup tools that MySQL can use, see the table below
Backup tools for MySQL
Backup method
Storage engine
Impact
Backup speed
Recovery speed
Recovery granularity
mysqldump
ALL
WARM
MEDUIM
SLOWEST
MOST FLEXIBLE
mysqldump
INNODB
HOT
MEDUIM
SLOWEST
MOST FLEXIBLE
select into outfile
ALL
WARM
SLOW
SLOW
MOST FLEXIBLE
mk-parallel-backup
ALL
WARM
MEDUIM
MEDUIM
FLEXIBLE
ibbackup
INNODB
HOT
FAST
FAST
FLEXIBLE
ibbackup
ALL
WARM
FAST
FAST
FLEXIBLE
backup command in mysqld
ALL
HOT
FAST
FAST
FLEXIBLE
filesystem (copy files)
ALL
COLD
FASTEST
FASTEST
NOT FLEXIBLE
snapshot (using LVM, ZFS, VMWare)
ALL
ALMOST HOT
FAST
FAST
LEAST FLEXIBLE
mysqlhotcopy
MyISAM
MOSTLY COLD
FAST
FAST
FLEXIBLE
The mysqldump program has been around a long time, it provides a logical backup of the entire database, individual databases, individual tables or even subsets of data using the --where option, it is often called a data dump. The output is in ascii format which means that you can open it in vi or notepad and change the contains if desired. I am not going to detail all options of the mysqldump command but show you a few examples
mysqldump ## backup all databases
mysqldump --user=root --password --all-databases > backup_<date>_all.sql

## backup a specific database
mysqldump --user=root --password <database_name> > backup_<date>_<database_name>.sql
## backup multiple databases
mysqldump --user=root --password <database_name>,<database_name> > backup_<date>.sql

## backup a table from a database
mysqldump --user=root --password <database_name> <table_name> > backup_<date>_<database_name>_<table_name>.sql
## backup some specific data
mysqldump --user=root --password <database_name> <table_name> --where "last_name='VALLE' order by first_name > backup_<date>.sql
## dumping from one database to another
mysqldump --databases <database_name> | mysql -h <destination_host> <database_name>
restore a mysqldump ## all databases
mysql --user=root --password < backup.sql

## specific database
mysql --user=<user> --password <database_name> < backup_<dataabse_name>.sql
You can use the into outfile clause of the select statement to backup individual tables, the command used to load the dump created is load data infile
select into outfile / load data infile ## dump of the accounts table
select * into outfile '/tmp/accounts.txt' from accounts;

## load the dump
load data infile '/tmp/accounts.txt' into table accounts;
The Maatkit parallel dump and restore toolkit can be downloaded from http://www.maatkit.org basically it's a wrapper around mysqldump which provides the programs mk-parallel-dump and mk-parallel-restore, what this means is that if you have a 16 core server and you are dumping 32 tables, the script will start up 16 separate copies of mysqldump with each process dumping a separate table.
mk-parallel-dump, mk-parallel-restore ## backup a database
mk-parallel-dump --basdir=/backups
## restore a database
mk-parallel-restore /backups
Snapshots for a filesystem depend on what operating system or software you are using, here are some links to my web pages regarding LVM, ZFS and VMWare
New in MySQL 5.6 is the online logical host backup, you can also use compression and encryption which is important when using sensitive data.
backup backup database <database_name> to '<database_name>-backup.sql'
restore restore from '<database_name>-backup.sql'
history select * from backup_history where backup_id = 321\G
There currently is a number of limitations of this command
  • no backup of the internal mysql datadisk
  • no native driver for InnoDB tables
  • no native driver for Maria or Falcon
  • no backup of partitions
  • no incremental backups
The mysqlhotcopy is a perl script written to provide a consistent backup of MyISAM and ARCHIVE tables, it does some limitations one of which when run it uses the lock tables command to create read locks on the tables being backed up, this allows for a consistent backup. again there are a number of options that you can use so have a look at the man page, here are a few examples
mysqlhotcopy ## backup a database
mysqlhotcopy <database_name> /backups

## backup multiple databases
mysqlhotcopy <database_name> accounts /backups

## backup a database to to another server
mysqlhotcopy --method=scp <database_name> \ username@backup.server:/backup

## use pattern match to backup databases and tables
mysqlhotcopy <database_name>./^employees/ /backup
Lastly ibbackup is a 3rd party software which allows you to perform non-blocking hot backups of InnoDB tables, it is entirely command-line driven which means that it is ideal for scripts, here is a link to the web site http://www.innodb.com/doc/hot_backup/manual.html
Recovering from Crashes
Most often you have to recover to a point-in-time after the last backup, the normal procedure is as follows
  • restore the latest backup
  • recovery the data to a point-in-time using recovery log files
MySQL server uses a binary format for the log files to save space, this means that you cannot view these files directly, a utility called mysqlbinlog is supplied to convert these log files into a text format that you can view. So the process for performing a point-in-time restore for MySQL is
  • restore the database using the last backup
  • determine the first binary log and starting position needed
  • determine the last binary log needed
  • convert the binary log to text format with the mysqlbinlog utility using options to specify the start and stop time
  • check the text file to make sure it's what you need
  • import the converted binary log(s)
convert the log files ## convert to a specific binary log file
mysqlbinlog mysql-bin.010310 > mysql-bin.010310.sql
## use a date to end at a specific time
mysqlbinlog --stop-datetime='201204-29 17:00:00' mysql-bin.010312 > mysql-bin.010312.sql

## other options are
--stop-datetime
--start-datatime
--start-position
--stop-position
restore the converted file mysql --user=root -password < mysql-bin.010310.sql