DEVELOPER ZONE ::
Login / Register
If you have followed the instructions, and your replication setup is not working, the first thing to do is check the error log for messages. Many users have lost time by not doing this soon enough after encountering problems.
If you cannot tell from the error log what the problem was, try the following techniques:
Verify that the master has binary logging enabled by issuing a SHOW MASTER STATUS statement. If logging is enabled, Position is non-zero. If binary logging is not enabled, verify that you are running the master with the --log-bin and --server-id options.
Verify that the slave is running. Use SHOW SLAVE STATUS to check whether the Slave_IO_Running and Slave_SQL_Running values are both Yes. If not, verify the options that were used when starting the slave server. For example, --skip-slave-start prevents the slave threads from starting until you issue a START SLAVE statement.
If the slave is running, check whether it established a connection to the master. Use SHOW PROCESSLIST, find the I/O and SQL threads and check their State column to see what they display. See Section 6.3, “Replication Implementation Details”. If the I/O thread state says Connecting to master, verify the privileges for the replication user on the master, the master hostname, your DNS setup, whether the master is actually running, and whether it is reachable from the slave.
If the slave was running previously but has stopped, the reason usually is that some statement that succeeded on the master failed on the slave. This should never happen if you have taken a proper snapshot of the master, and never modified the data on the slave outside of the slave thread. If the slave stops unexpectedly, it is a bug or you have encountered one of the known replication limitations described in Section 6.7, “Replication Features and Known Problems”. If it is a bug, see Section 6.12, “How to Report Replication Bugs or Problems”, for instructions on how to report it.
MySQL Enterprise. For immediate notification whenever a slave stops, subscribe to the MySQL Network Monitoring and Advisory Service. For more information see http://www.mysql.com/products/enterprise/advisors.html.
If a statement that succeeded on the master refuses to run on the slave, try the following procedure if it is not feasible to do a full database resynchronization by deleting the slave's databases and copying a new snapshot from the master:
Determine whether the affected table on the slave is different from the master table. Try to understand how this happened. Then make the slave's table identical to the master's and run START SLAVE.
If the preceding step does not work or does not apply, try to understand whether it would be safe to make the update manually (if needed) and then ignore the next statement from the master.
If you decide that you can skip the next statement from the master, issue the following statements:
mysql>SET GLOBAL SQL_SLAVE_SKIP_COUNTER =mysql>N;START SLAVE;
The value of N should be 1 if the next statement from the master does not use AUTO_INCREMENT or LAST_INSERT_ID(). Otherwise, the value should be 2. The reason for using a value of 2 for statements that use AUTO_INCREMENT or LAST_INSERT_ID() is that they take two events in the binary log of the master.
If you are sure that the slave started out perfectly synchronized with the master, and that no one has updated the tables involved outside of the slave thread, then presumably the discrepancy is the result of a bug. If you are running the most recent version of MySQL, please report the problem. If you are running an older version, try upgrading to the latest production release to determine whether the problem persists.
User Comments
Note that if you client does not do a "USE
dbname", binlog-do-db=dbname will not binlog a
query like: "update in dbname.foobar set foo=1"
You explicitly have to do a USE before a query in
order to have your query binlogged, it looks
like. Replication on the slave side can do
wildcard matches .. but the master cannot (a la
binlog-wild-do-table=dbname.%). So make sure your
clients do a use, if you plan to replicate those
tables it updates.
for MySQL v3.23.28:
When you attempt to use a certain master-
user/master-password combo to connect to the
mysql master, and you later change my.cnf to
attempt to connect with a new user, you must
update master.info to reflect the changes.
Since my master.info file only had one entry in
it (the slave only has one master), I simply
deleted the file. Upon restarting the slave
daemon, a new master.info was automatically
written.
--Curby
If you need to roll your own log file rotation
script as I did, and you are familiar with Perl,
look at the Log::Rotate module on CPAN rather
than reinventing the wheel.
If you intend to use "load table from master" you need to have access to that table to "rep" user, unlike this sections says of only having to set "File_priv" to "rep" user for all bases.
Above on this page where it talks about "properly
modularized and abstracted code" and refers to
safe_reader_query() and safe_writer_query(), I'd
like to put forth the proposition that different
abstracted functions for reading and writing need
not be nessessary for compliance with replication.
We currently have all of our queries running
through One safe-sql envelope, and we intend to
keep this architecture as we move to replication
by telling our envelope to send any queries that
begin with /\s+select/i to the slave, and
anything else to the master. We are running under
the assumption that all "read queries" are
selects. I can't think of any that aren't. If
there were they would probably be nominal in
performace draw (we use many complex select
statements :) and wouldn't do any harm being
handled by the master anyway.
So far as multi-statement queries, are code
doesn't, and won't, mix selects in the same query
with non-selects, or probably even use more than
one select per query, since we're not certain
what results could be returned in such a
circumstance. Thus reading and writing should
never get mixed up in the same query, and all
reads should start with the word "Select".
If you feel that my theory holds water then give
it a go, if you see a flaw in my logic before I
do mail me and lemme know, eh? 10x :)
- - Jess
When I started up my slave the first time, I had
been using binary logging for some time on the
master. I had already removed $hostname-bin.001
long ago. The slave complained about not being
able to find the first log and would not start
replaying transactions. To fix this, I stopped
the master, made a new snapshot, moved "*-bin.*"
to another directory, and started the master
again. Then when I put the snapshot on the slave
and started it, everything worked correctly.
If you want your slaves to connect to the
replication server with a unique username &
password and minimal privileges, you need to grant
just the FILE privilege to your replication user.
Similarly, if your slaves have no local updates
made on them, just lots of selects, its a very
good idea to connect with a user that cannot
update the data. This stops dead any chance of
mistakenly connecting to the wrong DB and losing
updates.
With verion 3.23.41 - I couldn't get replication
to work following the instructions about.
Specifically I got errors when I created
the "repl" user before copying data to the slave.
I could only get replication to work if I created
the "repl" user after copying data to the slave -
and obviously starting both servers...
Note: If you are running WinMySQLAdmin, you will
have to make the changes in the my.ini file as
well. It took me a while to realize this.
incredibly helpful page. I do have one issue to
raise - I have two servers doing 2-way
replication. One is on Linux, and one is on
Windows. There is an issue of case-sensitivity
in that if case is not taken into consideration
on the windows machine, the slave on the linux
machine stops. I'd love to hear any fixes to
this: [email protected]
I also had the problem Allen pointed out above,
in my case I had not actually deleted any of the
binary log files which the slave still required
to update from (I knew this courtesy of "show
slave status" command). But I had deleted several
log files it had finished with. I was able to fix
this without making a brand new snapshot (which
in my case took a larger outage that I would
like) What I did was edit the file (in the
Masters logs directory) $hostname-bin.index and
edited entries to show exact filenames of my
remaining binary log files.(ie: removing the
entries matching ones I had deleted manually from
the file system) I quickly did a mysql stop and
mysql start after that and I performed the
command "slave stop; slave start;" on the slave
and it started replicating again. I am running
version 3.23.41
Mysql 3.23.49
When a slave is running you have a master.info file. If you change the slave to become the master you must delete the master.info file or you may get errors such as this:
030130 10:57:03 Slave thread: error connecting to master: Unknown MySQL Server Host '' (4) (107), retry in 60 sec
It will keep retrying every 60 sec. Just deleting the master.info file worked for me.
I wouldn't suggest running replication on machines with different os. I had win2000 running MySQL server as well a RedHat Linux 7.3 with MySQL server as well. When the client code on Linux tried to connect to win2000 MySQL server, the server died, with win2000 reporting it has terminated.
Still haven't figure out what happened
Do NOT allow your replication master to run out of space in
the partition where you put your binlogs. One of two things
will happen:
1) the binlog will get truncated in the middle of an event, and the IO thread on the replication slave(s) will halt.
2) A series of transactions will go unrecorded on the master
binlog.
Either way, you'll have to reload the data from the replication master to the slave.
Having pulled my hair out with this error when getting a slave to replicate:
040308 13:38:01 Slave: reconnected to master '[email protected]:3306',replication resumed in log 'master-bin.011' at position 8379225
8
040308 13:38:01 Error reading packet from server: Access denied for user: '[email protected]' (Using password: YES) (server_errno=1045)
it turns out that I'd installed Mysql, then changed the hostname of the slave server - *however*, the slave server's mysql.user table still contained the old slave server's hostname.
Changing the host for the root user in the slave's mysql.user table to the correct hostname, then restarting the slave process fixed it all.
vittal
If you get a "Could not parse relay log event entry" error in a working replication setup and other slaves are fine, use STOP SLAVE; CHANGE MASTER TO master_log_file='(insert the value from Relay_Master_Log_File)', master_log_pos=(insert the value from Exec_master_log_pos); START SLAVE; . The slave will re-fetch from the master with no risk of losing a transaction. You should also look at the error log, in case it wasn't just a damaged network packet the first time.
SELinux can interfer with replication and be difficult to track down. This can sometimes occur after you have extracted files using TAR in the mysql data directory as is often instructed in replication examples because it can modify file properties tracked by SELinux. (Encountered using FC-4 SELinux and MySQL 4.1.11)
If you encounter the following error on the replication slave:
[ERROR] Slave I/O thread: error connecting to master '[email protected]:3306': Error: 'Can't connect to MySQL server on '192.168.7.11' (13)' errno: 2003 retry-time: 60 retries: 86400
check that you can login using:
mysql -u "replication_username" -h "replication_hostname" -p
if you can login using the above but replication fails to connect it may be SELinux policies restricting mysqld's access to ports, sockets, and files.
Check or grep /var/log/audit/audit.log and /var/log/messages for messages pertaining to mysqld like the following:
type=AVC msg=audit(1126806889.640:14244693): avc: denied { name_connect } for pid=2561 comm="mysqld" dest=3306 scontext=system_u:system_r:mysqld_t tcontext=system_u:object_r:mysqld_port_t tclass=tcp_socket type=SYSCALL msg=audit(1126806889.640:14244693): arch=40000003 syscall=102 success=no exit=-13 a0=3 a1=b139a130 a2=2 a3=0 items=0 pid=2561 auid=4294967295 uid=27 gid=27 euid=27 suid=27 fsuid=27 egid=27 sgid=27 fsgid=27 comm="mysqld" exe="/usr/libexec/mysqld"
To confirm SELinux is the cause, try turning off SE Linux from the root account using the following command:
# setenforce 0
# /etc/init.d/mysqld restart
If replication then starts working you know SELinux is the cause. On FC-4, running the following can shed light on this:
# audit2allow -i /var/log/audit/audit.log -l
allow mysqld_t mysqld_port_t:tcp_socket name_connect;
allow mysqld_t self:tcp_socket connect;
allow mysqld_t var_lib_t:dir { add_name read remove_name write };
allow mysqld_t var_lib_t:file { append create getattr lock read unlink write };
allow mysqld_t var_lib_t:sock_file { create getattr };
and running
# restorecon -R -v /var/lib/mysql
can fix problem associated with changing files in the MySQL directory
Since this post is about identifying the MySQL problems and not a tutorial about SELinux you should see also for details:
http://fedora.redhat.com/docs/selinux-apache-fc3/sn-debugging-and-customizing.html (which describes how to update the SELinux policies)
The master and slave must have the same value of global variable "collation_server", if not, slave io process can not be started.
If the slave was running previously but has stopped, the
reason usually is that some statement that succeeded on the
master failed on the slave. This should never happen if you
have taken a proper snapshot of the master, and never
modified the data on the slave outside of the slave thread.
AND you have never issued a statement on the MASTER which references a file or database that doesn't exist on the SLAVE.
For example:
INSERT INTO replicated_table SELECT stuff FROM master_only_db.table
Add your own comment.