You are browsing a read-only backup copy of Wikitech. The live site can be found at

MariaDB: Difference between revisions

From Wikitech-static
Jump to navigation Jump to search
(→‎Testing servers: Updated the info about the new testing hosts.)
Line 919: Line 919:

== Testing servers ==
== Testing servers ==
As of today (March 2020) there is one testing host on a shard called test-s4 that were bought for MCR testing
As of today (April 2021) there are two testing hosts on a shard called test-s4:

* <del>DB master: db1111 writable</del> (no longer available)
* DB master (writable): db1183.eqiad.wmnet  512GB RAM + RAID10 (same core specs)
* DB slave: db1077 read only (BBU is broken - which is not an issue for running queries). 512GB RAM + RAID10 (same core specs)
* DB slave: (read-only): db1125.eqiad.wmnet 512GB RAM + RAID10 (same core specs)

Whilst that hosts is normally up, it is not warranted it will always be available or up. It might have hardware issues (we have had to re-purpose some for production, and placed hosts with less HW reliability)
Whilst these hosts are normally up, it is not warranted they will always be available or up.  

That host have static data, and is not updated with production data. It has two wikis and a few tables (more tables can be requested if needed by filing a ticket to DBAs).
These hosts have static data, and is not updated with production data. More tables or wikis can be requested if needed by filing a ticket to DBAs).
It is not warranted that they'll be included, but they can be studied and some expectations about ETA and if it is possible to have that data will be given on the ticket.
It is not warranted that they'll be included, but they can be studied and some expectations about ETA and if it is possible to have that data will be given on the ticket.

Line 932: Line 932:

=== People with access ===
=== People with access ===
addshore [[phabricator:T211593|T211593]]
* Data Persistence Members
anomie [[phabricator:T194626|T194626]]
pmiazga [[phabricator:T189799|T219613]]
tarrow [[phabricator:T219613|T219613]]

== Stretch + 10.1 -> Buster + 10.4 known issues ==
== Stretch + 10.1 -> Buster + 10.4 known issues ==

Revision as of 11:18, 26 April 2021

For emergencies, check the subpage about MariaDB troubleshooting.

MariaDB is the main database management system used to run the Wikimedia sites.

For a general overview, check the MySQL@Wikipedia (2015) slides (MariaDB is a drop-in replacement for MySQL, which we used to use).

Sections and shards

For a full (?) list, look at db.php.

Core MediaWiki databases

  • s1: English Wikipedia
  • s2: 17 large wikis, including Chinese, Italian, Dutch, Polish, and Portuguese Wikipedias and English Wiktionary
  • s3: Most small wikis (~800)
  • s4: Commons
  • s5: German Wikipedia and 5 other wikis (T226950 proposes moving the largest wikis from s3 here)
  • s6: French, Japanese, and Russian Wikipedias
  • s7: CentralAuth, Meta, 10 large Wikipedias, and French Wiktionary
  • s8: Wikidata

External storage

The external storage servers host the full text of all page revisions in a compressed format.

  • es1: read-only cluster
  • es2: read-only cluster
  • es3: read-only cluster
  • es4 and es5: read/write cluster


Extension storage

  • x1: Notifications, Flow, ContentTranslation, Cognate, ReadingLists, UrlShortener. Tags for maintenance: DBA Operations Growth-Team Language-Team User-notice Cognate MediaWiki-extensions-UrlShortener StructuredDiscussions MediaWiki-extensions-BounceHandler ContentTranslation Reading List Service WikimediaEditorTasks Product-Infrastructure-Team-Backlog CommRel-Specialists-Support


The miscellaneous servers host databases for various auxiliary services.

Database creation template

If you need to request a new database please create a task with the DBA tag and filling out the following fields if you know them (or can estimate them)

  • QPS:
  • Size: (An estimation is fine)
  • DB Name:
  • User:
  • Accessed from server (s):
  • Backup Policy: Needed? Frequency?
  • Grants needed:



  • clouddb1021: Analytics dedicated replica. Used the first days of the month.
  • db1108: Eventlogging hosts
  • dbstore1003-1005: Multi-instance analytics hosts


The master database is the primary server for a given shard. If the master database server fails, then a replica can be promoted to be a new master. See Master switch for more information.

A database replica (or slave database) is another MariaDB or MySQL instance that replicates data from the master database. Most often, we add replicas in production for the purpose of load-balancing read queries, and for backup/failover purposes.

Cloud Services Wiki Replicas

Copies of the core databases with private data removed are part of the Data Services hosted by Wikimedia Cloud Services for bot, tool, and analysis use. The redaction is done during an intermediate stop on Sanitarium. As of Feb 2021, wikireplicas run in multiple hosts, with multiple processes, see new wiki replicas setup for more details.

To see how to operate (pool/depool) the new multi-instances check the new wiki replicas setup

Recloning a Wiki replica

This is an old procedure for the multi-source hosts, which are deprecated. This needs updating.

In case a Wiki Replica needs to be recloned from another one, there are several steps that needs to be done as these hosts are running multi-source.

  • Stop all slaves on the source and gather its replication coordinates
  • Stop MySQL on the source
  • Stop MySQL on the target host
  • Remove /srv/sqldata on the target host
  • Transfer the data from the source to the target: --no-encrypt --no-checksum SOURCE.eqiad.wmnet:/srv/sqldata TARGET.eqiad.wmnet:/srv
  • Make sure replication won't start automatically on neither of the hosts:
systemctl set-environment MYSQLD_OPTS="--skip-slave-start"
  • Once everything is transferred: on the target host remove the following files
  • Start MySQL on both hosts
  • If necessary run
  • On the target issue:
reset slave all;
  • On the target configure replication using the coordinates from the source that were gathered previously and start replication

Analytics MariaDB cluster

The Analytics MariaDB cluster contains full replicas of the core and extension storage databases.

Database backups

Main article: MariaDB/Backups

Replication protects against (in general) against hardware issues, as well as physical corruption. However, it is not enough to prevent application or operation errors. Also many people (e.g. developers and service maintainers) sometimes require observing past state of the database to solve data application issues.

In the past, 2 boxes replicate all shards using MariaDB 10 multi-source replication: dbstore1001 and dbstore1002 (as well as dbstore2001 and dbstore2002 on codfw). dbstore1001 and dbstore2001 also used to be delayed 24 hours. Multisource was abandoned for several reasons (task T159423): TokuDB (used for compression) bugs, difficult to recover and maintain, as well as difficulty for scalability as data grew more and more, MariaDB bugs and GTID limitations, among others.

In addition to the delayed slave, a bacula-orchestrated dumpdump was done weekly from dbstore1001 on misc and metadata core databases, and stored syncronously on bacula.

In 2017, multiple hardware and scalability issues lead to a focus and complete redesign of the database backup architecture, leading to the current one explained at MariaDB/Backups, with focus on full coverage, validation and testing and full automation recovery.

Start / Stop

In most cases, mariadb will not start automatically on server start. This is by design- a server is not ready to be put into production just after start for many reasons, plus it could be in a bad state (e.g. after crash, requiring upgrade or maintenance).

If a server has been shutdown normally, it can be started with:

systemctl start mariadb

(current MariaDB packages do not ship a systemd unit, although technically, jessie's backwards compatibility would make them work with systemctl. Avoid service, as it does not allow to pass extra parameters.

Right now, starting mysql does start the slave automatically- however, this again may not be desired if mysql_upgrade has to be run or the server otherwise checked. Prefer:

systemctl set-environment MYSQLD_OPTS="--skip-slave-start"
systemctl start mariadb

this will require later to run at the mysql prompt:




if it is using multi-source replication.

To shutdown, make sure mysql replication is not running first:

mysql> STOP SLAVE;

systemctl stop mariadb

If you just do "shutdown -[hr] now", there is a high chance that mysql will timeout and the operating system kills it uncleanly. This is usually due to one of these 3 reasons:

  • It takes a lot of time to empty the buffer pool (can take several minutes!)
  • The replication thread is killing/committing the ongoing transaction, and it takes a lot of time. This would seem that the slave is "blocked"
  • TokuDB has crashed and replication is "stuck" (in this last case, you actually have to kill the server)

If either action takes a long while:

journalctl -u mariadb

Do not assume that things will happen quickly, particularly if a box has been under heavy write load. Failing to wait for the server to stop will, with high probability, corrupt its data if non-transactional tables and GTID are being used- forcing to reload all data (a multi-hour or multi-day task!).

Consider also unmounting manually the /srv partition. Apparently, systemd does not respect the umounting time in some cases and it can lead to corruption:

puppet agent --disable "server shutdown" # prevents puppet messing up with the inexistent data
umount /srv

Buffer pool dump

In order to speed up the warming of the buffer pool (more frequent data should be in memory), automatic load/dump has been activated. See buffer pool dump for details.

InnoDB Shutdown Speed

InnoDB can take a while (hours) to write dirty pages out to disk on shutdown. If the instance must have a predictable downtime then make MariaDB begin flushing dirty pages well in advance of the shutdown process:

set global innodb_max_dirty_pages_pct=0;

Then poll:

show global status like '%dirty%';

When the counters approach zero, a shutdown and/or restart will be fast. Of course, since MariaDB is still handling traffic in the meantime, if write load is high the counters may never drop :-) Have to depool the box in that case.

Stopping a slave: avoiding replication corruption

Until GTID with transactional replication control is deployed, and probably forever on places like labs (where non-transactional tables are the norm), it is advised to stop replication first. In 99% percent of the cases, mysql already does that on shutdown, and shutdown does cleanly stop mysql -in theory. In reality, if shutdown takes too much time, systemd/rc/etc. kills mysql, which despite "server transactionality" (server is consistent on crash), that means that replication state is not itself transitional. This happens, for example, if long running updates (vslow, imports running on labs) prevent replication from stopping. In order to do that, just execute:

$ mysql --skip-ssl -e "STOP SLAVE"

This is specially true on crash: in most cases, reimaging a slave and reimporting it is the fastest way to assure data integrity (again, until production gets transactional replication state for InnoDB).


Asher started us using stock upstream packages from with the debs "de-debianized" each time to have fewer hooks, allowing puppet to handle stuff. Simply:

apt install wmf-mariadb101

We currently have wmf-mariadb10, wmf-mariadb101 (yes, I know, the names are horrible) and wmf-mysql80 built from source, with no patches except forcing the usage of openssl instead of yassl, that install to /opt.

Puppet controls manually the rc.d script and the package uses update-alternatives to update the symlinks in /usr/local/bin. Unlike the upstream versions our packages are fine to install alongside one another.


See mariadb roles in puppet.

Hardware installation checklist

Once the physical racking and OS installation has been, normally by DCOPs, the following items must be checked before considering the host ready to receive data.

  • Amount of memory showing up on the OS matches the expected
  • RAID setup
    • RAID level (typically 10)
    • RAID stripe size (256K)
    • BBU in place and policy set to WriteBack
  • Amount of disk space available for srv matches the expected
  • Number of CPUs matches the expected value

Setting up a fresh server

  • To initialize the database: /opt/wmf-mariadb104/scripts/mysql_install_db --basedir=/opt/wmf-mariadb104/
    • If this is a multi-instance db server, you'll need to provide --datadir=/srv/sqldata.<SECTION> as well.
  • Start the service: systemctl start mariadb
    • For multi-instance, the service is mariadb@<SECTION>
  • Set up defaults: /opt/wmf-mariadb104/bin/mysql_secure_installation --basedir=/opt/wmf-mariadb104/
    • For multi-instance, add -S /run/mysqld/mysqld.<SECTION>.sock
    • Don't set root password, accept defaults for everything else.

Loading Data / provisioning

Setting up a new replica (or repairing a current one) is done through the recovery/provisioning system: MariaDB/Backups

However, one can do still manual hot or cold copies of running or stopped servers through script on cumin hosts.

Schema Changes

First decide if the schema change can be done online. This only works if:

  • The table has a PRIMARY KEY or UNIQUE KEY on NOT NULL field(s)
  • MediaWiki can ignore the change behind a feature flag
  • Table metadata locking has been considered

Offline Schema Changes

  • Schedule downtime
  • See operations/mediawiki-config/db-eqiad.php
  • May need a master rotation?

Online Schema Changes

Use pt-online-schema-change or the ops wrapper script operations/software/dbtools/

$ --host=dbstore1001 --dblist=/a/common/s2.dblist --table=page_props \
  --method=percona "add pp_sortkey float default null"
  • --method=percona uses pt-online-schema-change and automatically checks slave lag on the host itself and its slaves.
  • --method=ddl uses normal ALTER TABLE which will probably cause replication lag.

The wrapper script prompts for confirmation of settings first and also does a dry run for each wiki. It generally requires little supervision but should never run unmonitored.

If modifying indexes, or adding new fields that are nullable and/or have a default value, the change can be applied to each slave first and their masters last, using --no-replicate; ie, alter the replication tree bottom-to-top so that no slave ever tries to replicate a non-existant field from a master. This:

  • Avoids risking the entire shard all at once; one can only break a single slave at a time :-)
  • Allows the slave to be depooled if something goes wrong or load is too high.
$ --host=dbstore1001 --dblist=/a/common/s2.dblist --table=page_props \
  --no-replicate --method=percona "add pp_sortkey float default null"

If the change simply must go to masters first and propagate to slaves through replication then the wrapper script may still be appropriate to use but such cases deserve careful consideration from a DBA. This especially includes any changes to primary keys!

Sanitarium and Labsdbs

See more: Sanitarium and labsdb

The sanitarium hosts have triggers defined on some tables which will clash with pt-online-schema-change. For small tables which should be fast to alter simply use --method=ddl. For large tables where DDL would cause unacceptable lag just schedule downtime.

Table Metadata Locking

Both ALTER TABLE and pt-online-schema-change need to hold the table metadata lock, even just briefly for the latter. They must wait for open transactions to close the table and also block new transactions opening the table. On a busy server and a hot table like page or revision this can easily result in many seconds of delay which is more than enough time for connections to pile up and hit max_connections.

Consider reducing load or depooling the box.

Replication Lag Trick

Issuing STOP SLAVE SQL_THREAD on a slave before starting pt_online_schema_change will cause immediate replication lag, which will in turn make MediaWiki load balancing code reduce traffic on the slave until it catches up. Just a few seconds of leeway is enough to allow the tool to create triggers, leading to far less interruption than waiting out the metadata locks traffic jam.


See MariaDB/monitoring.

Replication lag

Replication lag can be checked on

See MariaDB/troubleshooting#Replication lag and MySQL#Replication lag on how to handle replication lag.

Manipulating the Replication Tree

Tendril displays the full replication tree.

The following is interesting info, but thanks to gtid replication, implemented almost everywhere in production, except multi-source replicated slaves, you can move slaves just by executing:


A common task is manipulating slaves within the replication hierarchy. A script exists to help out: operations/software/dbtools/ It allows controlling two slaves in a parent/child or sibling/sibling relationship and will do sanity checks plus confirmation prompts. It has nothing to do with masters so don't try to use it for those :-)

Child to Sibling

Move a slave one layer upward in the hierarchy:

./ --switch-child-to-sibling --parent=db1007.eqiad.wmnet --child=db1035.eqiad.wmnet

Child must be replicating directly from parent.

Sibling to Child

Move a slave one layer downward in the hierarchy:

./ --switch-sibling-to-child --parent=db1007.eqiad.wmnet --child=db1035.eqiad.wmnet

Both slaves must be replicating from the same master.

Stop Siblings in Sync

Stop two slaves on the same layer in the hierarchy at a common binlog position:

./ --stop-siblings-in-sync --host1=db1007.eqiad.wmnet --host2=db1035.eqiad.wmnet

Both slaves must be replicating from the same master.

Multisource Slaves

A MariaDB 10 slave can have multiple masters. Have to set the default_master_connection session variable to indicate which stream is to be maniulated by subsequent commands:

./ --switch-child-to-sibling --parent=db1007.eqiad.wmnet --child=dbstore1001.eqiad.wmnet --child-set=default_master_connection=s7

Partitioned Tables

Links Tables

The *links tables tend to have data skewed toward certain namespaces depending on the wiki. In most cases this doesn't matter and the MariaDB optimizer always chooses a fast execution plan. However some edge cases on heavily used namespaces can cause massive filesorts. Historically MediaWiki used STRAIGHT_JOIN however that blunt instrument only introduced a different set of edge cases.

The *links tables respond well to range partitioning on namespace with ranges chosen appropriately on a case-by-case basis. Eg, commonswiki:

CREATE TABLE templatelinks (
  tl_from int(8) unsigned NOT NULL DEFAULT '0',
  tl_namespace int(11) NOT NULL DEFAULT '0',
  tl_title varbinary(255) NOT NULL DEFAULT ,
  UNIQUE KEY tl_from (tl_from,tl_namespace,tl_title),
  KEY tl_namespace (tl_namespace,tl_title,tl_from)
PARTITION BY RANGE (tl_namespace) (

Special Slaves

Shards S1-S7 have one slave each with table partitioning in place for revision and logging based on user id. These boxes handle special MediaWiki query groups like recentchangeslinked, contributions, and logpager.

Eg, from eswiki:

CREATE TABLE logging (
  log_id int(10) unsigned NOT NULL AUTO_INCREMENT,
  log_type varbinary(32) NOT NULL DEFAULT ,
  log_action varbinary(32) NOT NULL DEFAULT ,
  log_timestamp varbinary(14) NOT NULL DEFAULT '19700101000000',
  log_user int(10) unsigned NOT NULL DEFAULT '0',
  log_namespace int(11) NOT NULL DEFAULT '0',
  log_title varbinary(255) NOT NULL DEFAULT ,
  log_comment varbinary(255) NOT NULL DEFAULT ,
  log_params blob NOT NULL,
  log_deleted tinyint(3) unsigned NOT NULL DEFAULT '0',
  log_user_text varbinary(255) NOT NULL DEFAULT ,
  log_page int(10) unsigned DEFAULT NULL,
  PRIMARY KEY (log_id,log_user),
  KEY type_time (log_type,log_timestamp),
  KEY type_action (log_type,log_action,log_timestamp),
  KEY user_time (log_user,log_timestamp),
  KEY page_time (log_namespace,log_title,log_timestamp),
  KEY times (log_timestamp)
CREATE TABLE revision (
  rev_id int(8) unsigned NOT NULL AUTO_INCREMENT,
  rev_page int(8) unsigned NOT NULL DEFAULT '0',
  rev_text_id int(8) unsigned NOT NULL DEFAULT '0',
  rev_comment varbinary(255) NOT NULL,
  rev_user int(5) unsigned NOT NULL DEFAULT '0',
  rev_user_text varbinary(255) NOT NULL DEFAULT ,
  rev_timestamp varbinary(14) NOT NULL DEFAULT ,
  rev_minor_edit tinyint(1) unsigned NOT NULL DEFAULT '0',
  rev_deleted tinyint(1) unsigned NOT NULL DEFAULT '0',
  rev_len int(8) unsigned DEFAULT NULL,
  rev_parent_id int(8) unsigned DEFAULT NULL,
  rev_sha1 varbinary(32) NOT NULL DEFAULT ,
  PRIMARY KEY (rev_id,rev_user),
  KEY rev_timestamp (rev_timestamp),
  KEY page_timestamp (rev_page,rev_timestamp),
  KEY user_timestamp (rev_user,rev_timestamp),
  KEY usertext_timestamp (rev_user_text,rev_timestamp)

Queries that filter on log_user by equality or range run faster. Same for rev_user. The partition sizes are chosen on a per-wiki basis because one size does not fit all. Tendril has a report for choosing sizes based on row distribution.

Note the modified PRIMARY KEY definition that includes log_user. This is relatively safe for a slave but not appropriate for a master, so a partitioned slave should never be eligible for promotion. See coredb::$topology in puppet, or mediawiki-config/db-eqiad.php to identify them.

Extra notes: Having only 1 special node is a "Single Point of Slowdown". I am currently making sure that we have at least 2 nodes load-balancing this kind of traffic, which can be significant for uncached traffic, and allows to properly perform maintenance. Be careful, running the ALTER TABLE, as is, takes Query OK, 629385781 rows affected (5 days 54 min 6.51 sec) for the English wikipedia, and as much amount of space as the original table, so probably pt-online-schema-change or other methods can be considered to avoid failure and painful rollbacks.


Main page containing : HAProxy

dbproxy1XXX boxes run HAProxy. Besides making for easier Ops server rotations simply by having a proxy IP in the mix, there are two modes for other stuff: load balancing for slaves and failover for masters. So far, only misc shards masters and the latest labsdbs replicas uses it via dbproxy1XXX.


See role::mariadb::proxy::master

Load Balancing

See role::mariadb::proxy::slaves


The main module for DBAs on the operations/puppet tree is "mariadb", which is in its own repo operations/puppet/mariadb. Remember to update the subrepo when committing changes to the mariadb module, otherwise it will not be caught by palladium, strontium, CI, etc.

There used to be a class coredb_mysql, used from 5.5 nodes. Right now it is obsolete, but it is still in use by nodes that have not been updated to mariadb10, in particular, the masters. When all nodes are updated to 5.6/10, we will discontinue it, but right now it is essential for everything to work correctly.

Despite all the previous, there are mariadb-related files on the main repo- shared among the clases. Those are the puppet/templates/mariadb/ files, where the main configuration and grants lie.

There is a forth place where you will find mariadb-related files, and that is the private repository, but I suppose that requires no explanation (passwords, etc.).

Other DBA related repos are:

Long running queries

You can identify long running queries on tendril. If you need to kill a read query (e.g. causing too much lag, taking too long, etc.), log into the database host, enter MariaDB as root with sudo mysql and then run KILL <id>. Do not kill writes unless you are absolutely sure of what you are doing.

Older notes:

There is some event logic running on the servers trying to kill potential query exhaustion. Needs research.

I am investigating running:

pt-kill --print --kill --victims all --interval 10 --match-command Query --match-user wikiuser --group-by fingerprint --any-busy-time 50 --query-count 10 F=/root/.my.cnf

on a screen session to try to see if it is effective enough/does not create false positives. It requires more work, and finally, puppetization -then deciding which of the two options to follow.

Storage engines


In 2015, TokuDB was trialled but was found to be buggy. We're now back to using InnoDB.

Importing table spaces from other hosts with multi source replication

Transportable tablespaces

Since MySQL 5.6 it is posible to use advantage of transportable table spaces to move around .ibd files from one server to another (if file per table is enabled.

This feature provides a fast way of copying data from one host to another over the network using compression and nc for instance. To learn more about to copy .ibd files from one host to another, please check out MariaDB/ImportTableSpace.

Account handling

root@localhost account no longer uses passwords for authentication (task T150446), but the UNIX socket authentication plugin. This allows stop using passwords (that can be copied, exposed and compromised), and allows the root system user to login from localhost by running:

sudo mysql


sudo mysql --skip-ssl

Because mariadb uses the uid of the linux user, there is no need to write passwords to the filesystem anymore, and in the event of a host compromise, only localhost is affected, and not any other host sharing the same password. This could later be extended to other system accounts that only connect from localhost, such as icinga or prometheus. Note that if a superuser is compromised on a host, not having a password is not a further barrier, as root has the ability to manage, read and modify at will MySQL files.

How unix_socket authentication works

All mysql servers should have been configured to load the auth_socket plugin (

plugin-load =

It can also be enabled at runtime with:

INSTALL PLUGIN unix_socket SONAME 'auth_socket';

That only enables to plugin, to use it, we have to alter the user we want to authenticate:

GRANT USAGE ON *.* TO root@localhost IDENTIFIED VIA unix_socket;

This will delete the user password, but will indicate it does not uses the mysql native password authentication, but unix socket.

MariaDB [(none)]> SELECT user,host,password,plugin FROM mysql.user WHERE user='root';
| user | host         | password                                  | plugin      |
| root | localhost    |                                           | unix_socket |

By the way, note that the authentication name is unix_socket, but the plugin loaded is Do not use auth_socket on the GRANT/CREATE USER statements.

Also, if for some reason, you revert this change, make sure you put a password back:

GRANT USAGE ON *.* TO root@localhost IDENTIFIED BY '<super_secure_password>';

Dumping tables with mydumper

What is mydumper?

From the documentation:

* Parallelism (hence, speed) and performance (avoids expensive character set conversion routines, efficient code overall)
* Easier to manage output (separate files for tables, dump metadata, etc, easy to view/parse data)
* Consistency - maintains snapshot across all threads, provides accurate master and slave log positions, etc
* Manageability - supports PCRE for specifying database and tables inclusions and exclusions

Why are we starting to consider it?

Its speed and parallelism makes it perfect to save time and load data a lot faster.

It is easier to recover single tables or rows.

It compress nicely

From the tests we have seen

  • Taking a full dump of s3 takes 1:15h - 88GB
  • Taking a full dump of s1 takes 53 minutes - 93GB
  • Taking a full dump of s5 takes 1:20h - 93G

Quick cheatsheet

Dump data

Right now and in order to start it - as it doesn't accept any flag as no to read the default file, the following section needs to be commented out on the host my.cnf

# ssl-ca=/etc/ssl/certs/Puppet_Internal_CA.pem
# ssl-cert=/etc/mysql/ssl/cert.pem
# ssl-key=/etc/mysql/ssl/server.key
# ssl-verify-server-cert

Once that is done, a typical way to dump a database:

mydumper -c -h localhost -t 8 -u root -r 100000000 -B wikidatawiki -S /tmp/mysql.sock -o output_directory_name

-c To compress the data files

-h The host

-t How many parallel threads you want dumping at the same time

-u The user

-r Try to split tables into chunks of this many rows

-B Database you want to dump

-S Specify the socket to connect to

"-o" Specify the output directory where you want to save the files to. If not specified, it will generate a directory under your current path called: export-2017xxxxx-xxxxx with all the files.

Load data

Please check the dump data section to make sure you are able to connect (comment out the SSL options).

By default it disables the logging into the binlog

Once that is done, typically you want to run:

myloader -h localhost -t 8 -u root -S /tmp/mysql.sock -d export-xxxxx

-h The host

-t How many parallel threads you want loading the data at the same time

-u The user

-S Specify the socket to connect to

-d The directory you want to import the files from

Production section failover checklist

It has been suggested that this page should be merged with MariaDB/troubleshooting#Depooling_a_master_(a.k.a._promoting_a_new_slave_to_master)

Note there is a script. db-switchover that automates most of these steps.

NEW master: OLD master:

  1. Check configuration differences between new and old master
pt-config-diff h=NEW.eqiad.wmnet,F=/root/.my.cnf h=OLD.eqiad.wmnet,F=/root/.my.cnf 

  1. Silence alerts on all hosts
  2. Set NEW master with weight 0 and depool it from its section
dbctl instance NEW edit
dbctl config commit -m "Set NEW with weight 0 TXXXXXX"
  1. Topology changes, connect everything to NEW
db-switchover --timeout=15 --only-slave-move OLD.eqiad.wmnet NEW.eqiad.wmnet
  1. Disable puppet @NEW and @OLD
 puppet agent --disable "switchover to NEW"
  1. Merge gerrit puppet change to promote NEW to master (Example):
  1. Start the failover
!log Starting sX failover from OLD to NEW - TXXXXXX
dbctl --scope eqiad section sX ro "Maintenance till 05:30AM UTC TXXXXXX" && dbctl config commit -m "Set sX as read-only for maintenance TXXXXXX"
  1. Check that sX is indeed on read-only
  2. run switchover script from cumin1001:
root@cumin1001:~/wmfmariadbpy/wmfmariadbpy# db-switchover --skip-slave-move OLD NEW ; echo OLD; -hOLD -e "show slave status\G" ; echo NEW ; -hNEW -e "show slave status\G"
  1. Promote NEW to master and remove read-only, leave OLD (old master) with weight 0 for now
dbctl --scope eqiad section sX set-master NEW && dbctl --scope eqiad section sX rw && dbctl config commit -m "Promote NEW to sX master and remove read-only from sX TXXXXXX"
  1. Restart puppet on old and new masters (for heartbeat): OLD and NEW
 run-puppet-agent -e "switchover to NEW"

(Equivalent manual steps)

  1. Set old master in read only: -hOLD -e "set global read_only=ON;"
  2. Disable heartbeat @OLD
 killall perl
  1. Confirm new master has catched up
 host=OLD; echo "== $host =="; -h $host -e "show master status\G show slave status\G" | grep ' File:\| Position:\|Relay_Master_Log_File:\|Exec\_Master\_Log\_Pos:'; host=NEW; echo "== $host =="; -h $host -e "select @@hostname; show master status\G show slave status\G" | grep ' File:\| Position:\|Relay_Master_Log_File:\|Exec\_Master\_Log\_Pos:'
  1. Stop slave on new master -h NEW -e "STOP SLAVE;" 
  1. Confirm pt-hearbeat has been started on NEW
 pgrep perl  / -h NEW -e "SELECT * FROM heartbeat.heartbeat ORDER BY ts DESC LIMIT 1\G"
  1. Remove read-only from NEW: -hNEW -e "set global read_only=OFF;"
  2. RESET SLAVE ALL on new master -h NEW.eqiad.wmnet -e "reset slave all;"
  1. Change old master to replicate from new master -hOLD: change master to master_host='NEW.eqiad.wmnet', master_user='repl', master_password=, master_port=3306, master_log_file=, master_log_pos=, master_ssl=1;

if all looks good: start slave; Enable GTID on old master (OLD)

(End of equivalent manual steps)

Clean up tasks:

  1. change events for query killer:
   events_coredb_master.sql on the new master NEW
   events_coredb_slave.sql on the new slave OLD
  1. Update DNS (example):
  2. Give weight to OLD if needed
  3. If the old master will be decommissioned or will go under maintenance: depool it from dbctl: dbctl instance HOSTNAME depool
  4. Change candidate master note (generally remove it from NEW and add it to OLD)
dbctl instance OLD set-candidate-master --section sX true
dbctl instance NEW set-candidate-master --section sX false
  1. Update/resolve phabricator ticket about failover

External store section failover checklist

External store failovers differ a bit from current sX ones, as we need to disable writes on the given section first. RO external stores (as of today, es1, es2 and es3) do not need this, as those hosts are stand alone and only read only.

As of today, es4 and es5 are RW, so they have a normal replication topology.

Disabling writes on a section can be done safely, so the failover needs no rush. Even if it can be done hours ahead, disabling writes should be done just a few minutes the actual failover time, to avoid un-balancing of both sections.

When one section has writes disabled, reads are still happening without any issues.

If the idea is to depool the old master, make sure to leave the new master with a bit of weight, to avoid having just one slave serving traffic. Usually leaving the master with weight 50 and the slave with weight 100 is enough.

Check list for failing over an es section with es1023 and es1024 as examples:

NEW master: es1024
OLD master: es1023
# Check configuration differences between new and old master
$ pt-config-diff h=es1023.eqiad.wmnet,F=/root/.my.cnf h=es1024.eqiad.wmnet,F=/root/.my.cnf  

# Silence alerts on all hosts
# Set NEW master with weight 50
dbctl instance es1024 edit
dbctl config commit -m "Set es1024 with weight 50 T255755"
# Topology changes, connect everything to es1024
db-switchover --timeout=15 --only-slave-move es1023.eqiad.wmnet es1024.eqiad.wmnet
# Disable puppet @es1023 and @es1024
 puppet agent --disable "switchover to es1024"
# Merge gerrit puppet change to promote es1024

# Start the failover  
!log "Starting es failover from es1023 to es1024 -"
Disable es5 writes
# Check that es5 is indeed on read-only
# run switchover script from cumin1001:
root@cumin1001:~/wmfmariadbpy/wmfmariadbpy# db-switchover --skip-slave-move es1023 es1024 ; echo es1023; -hes1023 -e "show slave status\G" ; echo es1024 ; -hes1024 -e "show slave status\G"
# Promote es1024 to master and remove read-only, leave es1023 (old master) with weight 0  
dbctl --scope eqiad section es5 set-master es1024 && dbctl config commit -m "Promote es1024 to es5 master T255755"
# Restart puppet on old and new masters (for heartbeat): es1023 and es1024
 run-puppet-agent -e "switchover to es1024"
# Enable es5 on MW (REVERT PATCH

Clean up tasks:
# change events for query killer:
   events_coredb_master.sql on the new master es1024
   events_coredb_slave.sql on the new slave es1023 
# Update DNS:
# Update/resolve phabricator ticket about failover

Misc section failover checklist (example with m2)

OLD MASTER: db1065

NEW MASTER: db1132

  1. Check configuration differences between new and old master
$ pt-config-diff h=db1068.eqiad.wmnet,F=/root/.my.cnf h=db1081.eqiad.wmnet,F=/root/.my.cnf 
  1. Silence alerts on all hosts
  2. Topology changes: move everything under db1132
db-switchover --timeout=1--only-slave-move db1065.eqiad.wmnet db1132.eqiad.wmnet
  1. Disable puppet @db1065, puppet @db1132
 puppet agent --disable "switchover to db1132"
  1. Merge gerrit:
  2. Run puppet on dbproxy1002 and dbproxy1007 and check the config
puppet agent -tv && cat /etc/haproxy/conf.d/db-master.cfg
  1. Start the failover
!log Failover m2 from db1065 to db1132 - T226952
 root@cumin1001:~/wmfmariadbpy/wmfmariadbpy# db-switchover --skip-slave-move db1065 db1132 
  1. Reload haproxies
dbproxy1002:   systemctl reload haproxy && echo "show stat" | socat /run/haproxy/haproxy.sock stdio
dbproxy1007:   systemctl reload haproxy && echo "show stat" | socat /run/haproxy/haproxy.sock stdio
  1. kill connections on the old master (db1065)
 pt-kill --print --kill --victims all --match-all F=/dev/null,S=/run/mysqld/mysql.sock
# START Equivalent manual steps
# Disable GTID on db1132
# Disable heartbeat @db1065
  killall perl
# Set old m2 master in read only
  ./ -h db1065 -e "SET GLOBAL read_only=1"
# Confirm new master has catched up DONE
  host=db1065 echo "== $host =="; ./ -h $host -e "show master status\G show slave status\G" | grep ' File:\| Position:\|Relay_Master_Log_File:\|Exec\_Master\_Log\_Pos:'; host=db1132; echo "== $host =="; ./ -h $host -e "select @@hostname; show master status\G show slave status\G" | grep ' File:\| Position:\|Relay_Master_Log_File:\|Exec\_Master\_Log\_Pos:'
 # Reload haproxies
 dbproxy1002:   systemctl reload haproxy && echo "show stat" | socat /run/haproxy/haproxy.sock stdio
 dbproxy1007:   systemctl reload haproxy && echo "show stat" | socat /run/haproxy/haproxy.sock stdio
 # Restart puppet on old and new masters (for heartbeat): db1065 and db1132
  puppet agent --enable && puppet agent -tv
# Set new master as read-write and stop slave
  mysql -h db1132.eqiad.wmnet -e "STOP SLAVE; SET GLOBAL read_only=0;"
# RESET SLAVE ALL on new master
  ./ -h db1132.eqiad.wmnet -e "reset slave all;"
# Change old master to replicate from new master DONE
  ./ -hdb1065: change master to master_host='db1132.eqiad.wmnet', master_user='repl', master_password=, master_port=3306, master_log_file='xx', master_log_pos=xx, master_ssl=1;
if all looks good: start slave; 
# Update tendril and zarcillo master server id for m2 
  UPDATE shards set master_id = 1628 where name='m2' LIMIT 1; -h db1115 zarcillo -e "UPDATE masters SET instance = 'db1132' WHERE dc = 'eqiad' and section = 'm2' LIMIT 1" 
# Enable GTID on the old master db1065
# STOP equivalent manual steps

  1. Restart puppet on old and new masters (for heartbeat):db1065 and db1132
 puppet agent --enable && puppet agent -tv
  1. Check services affected (otrs,debmonitor) DEBMONITOR and OTRS looking good
  2. Clean orchestrator heartbeat to remove the old masters' one, otherwise Orchestrator will show lag
  3. change events for query killer:
   events_coredb_master.sql on the new master db1132
   events_coredb_slave.sql on the new slave db1065 
  1. Update/resolve phabricator ticket about failover
  2. Create decommissioning ticket for db1065

Special section: x1 master switchover

x1 is a "special" section which cannot be put on read-only on mediawiki, so it needs to relay on the db-switchover script which puts MySQL on read-only. When failing over, please tag the following Teams and people on the phabricator task so they can have a heads up as they are x1 stakeholders and need to know that x1 will have writes blocked for around one minute.

  • Subscribers: Tgr, JoeWalsh, Dbrant, Ladsgroup, Addshore, Legoktm, Mholloway
  • Tags: Cognate, Growth-Team, Language-Team, User-notice, UrlShortener, StructuredDiscussions, MediaWiki-extensions-BounceHandler, ContentTranslation, Reading List Service, WikimediaEditorTasks.

Example task:

Decommissioning a mariadb host server checklist

See MariaDB/Decommissioning_a_DB_Host

Depool a broken or lagged replica

From cumin1001 or cumin2001:

dbctl instance dbXXXX depool
dbctl config commit

Create a task with the DBA tag so DBAs can follow up and checkout what happened, a proper fix etc

THE SECTION BELOW IS DEPRECATED: Checkout the mediawiki-config repo if you haven't already:

git clone ssh://
cd mediawiki-config
vim wmf-config/db-eqiad.php

Assuming eqiad is the active DC

  • A replica can run just one MySQL instance
  • If a replica runs more than one MySQL instance, it will be specified in its definition with the following format



This means db1090 has an instance running on 3317 (and probably others in different ports). The port is formed in the following way

3306 -> MySQL standard port
331X -> X meaning the section where the server belongs to. 
1 -> s1
2 -> s2
8 -> s8
3320 -> x1
3321 -> m1
3322 -> m2
3323 -> m3
3324 -> m4
3325 -> m5

So db1090:3317 is a replica that runs on s7

  • A replica can handle more than one type of traffic.

These are the types of traffic we have


They are normally specified on the definition files, let's use db-eqiad.php as it is the active datacenter Let's examine s7 where db1090 belongs to: This is the main traffic section where the servers and weights are specified. Servers with weight 1 normally means they are still checked for lag by the MediaWiki LB, but the have almost no main traffic, because they are special replicas and they have more weight for special traffic (check below)

's7' => [
'db1062' => 0,      # D4 2.8TB 128GB, master
'db1079' => 300,    # A2 3.6TB 512GB, api # master for sanitarium db1125
'db1086' => 400,    # B3 3.6TB 512GB, api # candidate master
'db1090:3317' => 1, # C3 3.6TB 512GB, vslow, dump: s2 and s7, old master
'db1094' => 500,    # D2 3.6TB 512GB
'db1098:3317' => 1, # B5 3.6TB 512GB # rc, log: s6 and s7
'db1101:3317' => 1, # C2 3.6TB 512GB # rc, log: s7 and s8

And there is also a special slaves section below, look for it: db1090:3317 is a single vslow slave db1079 and db1086 share API traffic, db1079 has more weight for API because db1086 has more weight for main traffic (see above) db1098:3317 db1101:3317 are recentchanges replicas and they share the same amount of load for all those special services.

's7' => [
 'vslow' => [
  'db1090:3317' => 1,
 'dump' => [
  'db1090:3317' => 1,
 'api' => [
  'db1079' => 3,
  'db1086' => 1,
 'watchlist' => [
  'db1098:3317' => 1,
  'db1101:3317' => 1,
 'recentchanges' => [
  'db1098:3317' => 1,
  'db1101:3317' => 1,
 'recentchangeslinked' => [
  'db1098:3317' => 1,
  'db1101:3317' => 1,
 'contributions' => [
  'db1098:3317' => 1,
  'db1101:3317' => 1,
 'logpager' => [
  'db1098:3317' => 1,
  'db1101:3317' => 1,

Locate the broken slave - keep in mind it can be in more than one section. The easiest way is probably to look for it on the file. For this example let's assume db1090:3317 is broken with replication broken, but its other instance (3312) is working fine, so no need to depool it.

This host is a vslow host, so we need to replace it with another host from the same section. Let's chose db1079 to replace it.

If the host is a special host and servers some special traffic (vslow,api, recentchanges, it needs to be replaced if it is the only one, we cannot leave a section without a slave)

We need to look for it and comment it out from every single line where db1090:3317 appears. However DO NOT COMMENT the line with its definition which is already really explicit:

    'db1090:3317' => '', # do not remove or comment out

So this should be the diff once we are ready to commit:

git add wmf-config/db-eqiad.php
git commit
git review

Once we have +2 on gerrit and the change has been merged automatically:

ssh deployment.eqiad.wmnet
cd /srv/mediawiki-staging/
git fetch ; git rebase
git show (this should show our change)
scap sync-file wmf-config/db-eqiad.php "Depool db1090:3317 as it is broken"

Once the change is deployed, we should be able to see our change on:

If a host is completely down and it has multiple instances it needs to be depooled from all the instances Let's say db1090 had a hardware crash and it is powered off, that means that both instances 3312 and 3317 are down. We need to also depool the instance db1090:3312 with the same method as described above.

Create a task with the DBA tag so DBAs can follow up and checkout what happened, a proper fix etc

Testing servers

As of today (April 2021) there are two testing hosts on a shard called test-s4:

  • DB master (writable): db1183.eqiad.wmnet 512GB RAM + RAID10 (same core specs)
  • DB slave: (read-only): db1125.eqiad.wmnet 512GB RAM + RAID10 (same core specs)

Whilst these hosts are normally up, it is not warranted they will always be available or up.

These hosts have static data, and is not updated with production data. More tables or wikis can be requested if needed by filing a ticket to DBAs). It is not warranted that they'll be included, but they can be studied and some expectations about ETA and if it is possible to have that data will be given on the ticket.

Access to that hosts is granted individually to those who already have an NDA and cluster access if _really_ needed. Please file a ticket and tag DBA to let us know why you need access so we can start a discussion about it.

People with access

  • Data Persistence Members

Stretch + 10.1 -> Buster + 10.4 known issues

  • Fixed: Optimizer flag rowid_filter needs to be disabled until it is fixed upstream (and checked) as it causes a regression: T245489 Upstream:
  • Possible bug with the optimizer choosing the wrong query plan on certain queries, pending upstream pick up: T246069 Upstream:
  • Some metrics have been deprecated in 10.4 and need to be removed from grafana dashboards: T244696
  • prometheus-mysqld-exporter:
  • wmf-pt-kill seems to be working fine, but needs more checking on wiki replicas.
  • Tendril doesn't like this in-place upgrade, so it requires a disable + drop + add + enable after upgrade, otherwise the Act. (last contact) field doesn't get updated.
  • Fixed: Events on 10.4 starts disabled: T247728 - to be fixed on 10.4.13
  • On mysql_upgrade, backup user gets an additional grant (as it has SUPER, needed to handle replication) "DELETE HISTORY". This makes mydumper fail with "broken table detected" on sys and other system tables (my guess is because it can list it, but cannot read from it - this is on purpose, we only want backups from certain databases). Revoking the grant fixed mydumper run:
REVOKE DELETE HISTORY ON *.* from 'dump'@'X.X.X.X';

How to enable a new external storage (es) section

The process of enabling a new external storage section can be complex and require different steps.

Pre steps

  • Check all the involved servers have replication working and enabled (with SSL). Check replication lag is 0
  • Check all the involved servers have weight (if not done, check the section below)
  • Check pt-heartbeat is running
  • Check all the involved servers have notifications enabled (and Icinga looks clean)
  • Check the future active master has read_only set to OFF
  • Check the future servers have all the wikis and blobs_clusterXX tables created. This is an example ticket T245720
  • Check wikiadmin wikiuser have grants.

Setting up the servers with dbctl

external storage sections normally have the following weights: master: 0 slave1: 100 slave2: 100

In order to be able to do so, use dbctl to generate those configs.

dbctl instance esXXXX edit

And leave it like this for the master

# Editing object eqiad/esXXXX
host_ip: 10.XX.XX.XX
port: 3306
  esX: {percentage: 100, pooled: true, weight: 0}

And like this for a slave

# Editing object eqiad/esXXXX
host_ip: 10.XX.XX.XX
port: 3306
 esX: {percentage: 100, pooled: true, weight: 100}

Commit the change

dbtcl config commit -m "Set weights for the new es hosts - TXXXXXX"

Mediawiki patches and deployment

In order to be able to set the new external storage there are two steps needed from mediawiki point of view. They can be just one, but in order to catch errors, it better be split into two.

Once the change is merged and deployed to the passive DC, use Wikimedia Debug extension to browse the site via mwdebug2001.codfw.wmnet and monitor errors. If everything looks fine, deploy on the active DC.

Use shell.php from a mwmaint host to fetch stuff from those new tables, to make sure grants and everything is fine. Getting a false is a good thing. Everything else might require further research.

root@mwmaint1002:~# mwscript shell.php --wiki=enwiki
Psy Shell v0.9.12 (PHP 7.2.26-1+0~20191218.33+debian9~1.gbpb5a340+wmf1 — cli) by Justin Hileman
>>> ExternalStore::fetchFromURL( 'DB://cluster26/1' )
Wikimedia/Rdbms/DBConnectionError with message 'Cannot access the database: No working replica DB server: Unknown error'
>>> ExternalStore::fetchFromURL( 'DB://cluster25/1' )
=> b"""

If everything looks fine, the next patch should be enabling that section to be writable.

Before fully deploying that change, test it on mwdebug. Generate some writes by writing on your talk page and checking on the new blobs_clusterXX table. It might take a few writes to be able to show up there, as it is balanced with the other existing external storage masters.

Try a few times, and a few projects to make sure the row shows up there and on all the hosts (to double check replication is working as expected).

If the writes are showing up, and nothing else is errorring, you are good to go and fully deploy to all DCs.

Once deployed, monitor for lag, errors and periodically select from the new tables to make sure it is getting new things.

How to set a external storage (es) section to read only

Whenever an existing external storage section needs to go read-only, these are the steps that need to be taken.

As always, first push to the passive DC, monitor for logs and if all is ok, push to the active DC:

  • Once the change is live, monitor the master binlog (skip heartbeat entries) and make sure the INSERTS stopped a few seconds after the change is live everywhere.
  • Disable alerts for 10-15 minutes on the involved servers, to avoid any alerts.
  • Once writes has stopped, prepare and push a puppet patch to change those hosts to standalone:
  • Set read_only=1 on the master
  • Once puppet has run on the master, check that pt-heartbeat has stopped there. Once done, run show master status; to make sure nothing is being written.
  • Run the following across all the hosts: show master status; show slave status\G and note those positions somewhere.
  • Reset replication across all the hosts: stop slave; reset slave all
  • Update tendril to hide that replication tree:
update shards set display=0 where name='esX'
  • Update zarcillo to reflect those hosts as standalone
 update sections set standalone=1 where name='esX';
  • Slowly give some weight to the old masters, and now standalone - probably best done in a few commits, to slowly warm their buffer pool
dbctl instance esXXXX edit
dbctl config commit -m "Slowly give weight to old and now standalone es masters TXXXXXX"

dbctl at the moment doesn't treat standalone hosts any different from a normal replicating section. So it will be showed normally on either db-eqiad.json and db-codfw.json even though there is no real master anymore but just standalone hosts.

IPv6 and MariaDB

MariaDB (and MySQL) should have good support of IPv6. However, due to account handling being based on IP addresses, not DNS, a lot of issues, specially with authentication and grants can arise:

  • Hosts using DNS will try to contact mysqld using IPv6 first. That will cause timeouts (of several seconds) until it fails back to IPv4
  • Privileges may fail completely (access denied):
    • When setting up and performing mysql backups
    • When accessing host from remote mysql root clients (e.g. cumin)
    • When using replication
    • When using watchdog user for tendril

db1108, as of July 2020, is the only host configured using IPv6 while holding a mysql service.

This page is a part of the SRE Data Persistence technical documentation
(go here for a list of all our pages)