The root cause of this issue is that Red Hat broke binary compatibility of their OpenSSL packages between 6.4 and 6.5, something that they promised that they would not do.
Resolving this problem is straightforward, but depending on the applications you may have deployed, may take some yelling at your application vendors. Be sure to reserve most of your ire ...
No because the above mentioned --apply-log won't update LSN in xtrabackup_checkpoints, thus every next incremental backup will copy pages modified since the last full backup. That's not what you want to achieve
To implement your scenario you need:
Take full backup
innobackupex --no-timestamp /path/full
Save the last LSN
# cat /path/full/...
You can install PS 5.6 once you upgraded openssl to openssl-1.0.1e-15.el6.x86_64.rpm
For 6.4 we (I work for Percona) also have some custom made packages available:
This is a bug of Percona Server 5.7:
This was introduced as part of our work in porting Facebook MySQL
MyRocks engine to Percona Server 5.7. The use of I_S over P_S was not
caught during the merge of this feature from 5.6 to 5.7
An immediate workaround is to change the show_compatibility_56 back to
the default value of ON which will expose the ...
Postfix depends on having a package that provides libmysqlclient.
You can remove the old packages (mysql57-community) and install the new ones (Percona) in a single yum transaction using yum swap. This will ensure that Postfix will have the dependency before and after the transaction and yum will not try to remove it.
yum swap -- install percona-whatever-...
The first point to remember here is many mysql clients (including php) will default to unix socket style communication if the client is configured to use localhost. If your server is not correctly setting up this socket, then clients may be attempting to connect to the socket and failing.
The rest is summarized from our discussion in the comments.
In any ...
From the show status output your servers don't load any wsrep provider (Galera), and probably don't even attempt to.
As little as I understand in how mysql configuration file works, I think you need to put [mysqld] before MariaDB/Galera settings (or put them immediately after InnoDB settings). Otherwise they seem to be in the [isamcheck] section and simply ...
What if you change this part in my.cnf
I guess systemd did kill MySQL because it did not find a proper pid file containing a process number in that location.
Answer to Question 1
pt-table-sync (togerther with pt-table-checksum) can be used to correct replication errors like data corruption, someone directly modifying data on the slave, server crashes, schema changes in wrong order, etc.
However pt-table-sync can be also used without replication for keeping tables synced in near-realtime if the data don't change ...
No, you can't run it on another machine, as it reads the data files directly, not via the mysql server process.
However, you can mount disk space of another machine on the database host and use this space as the target directory for the backup.
I've also suffered from a problem with similar symptoms (all nodes, VMs, and CTs go into an "unknown" status). Using the command line everything seemed fine and so it was more of a nuisance than anything because it meant I had to migrate everything and reboot each node individually before I could use the web UI again. I eventually figured out that restarting ...
You need to install Percona Xtrabackup on the remote server and then SSH from your local server to the remote server. From this position, you can initiate the backup and stream to your local server.
Xtrabackup requires access to MySQL's data directory (and MySQL instance as well), so it can't be run on another server. However, it can be triggered by ...
I don't think this is a memory leak. It could simply be that without large pages your memory usage hit an upper bounds (ceiling) which was why memory usage looks flat. After switching to large pages MySQL may have simply been able to better make use of the available memory. The system isn't hitting swap. I'd say closely monitor and stop flushing tables.
It turned out that mysql wasn't the culprit. In php-fpm request_slowlog_timeout was set to 1s. Increasing this value fixed the errors. It seems that fpm was killing processes because the timeout increased 1 second.
I've did this , exactly this
... since the results with XtraDB ON TOP of gluster were mixed ,
in the end the thingy was moved to a triple-headed percona Cluster with ProxySQL , your database speeds will be much faster without the "high-latency" glusterFS Layer
(at least in production database scenario with "live responses" , if ...
Replication is for availability,you might be interested in things like pacemaker clustering and the Percona mysql resource agent to manage replication and provide automatic failover.
Considering innobackupex, it is primary a backup utility, replication might go wrong, in that case you'll be happy to have a fast way of recovery.
Hope this helps.
It looks like the packages provided by remi aren't compatible with the ones supplied by Percona. Fortunately it looks like Percona supply the packages too
yum list | grep percona | grep devel
Percona-SQL-devel-50.x86_64 5.0.92-b23.89.rhel6 percona
Percona-Server-devel-51.x86_64 5.1.73-rel14.11.603.rhel6 percona
Table alteration involves /tmp partition as intermediate storage for explicitly and implicitly created temporary tables. You have to use the tmpdir on a partition that big enough. Add the next line to the my.cnf
tmpdir = /path/to/the/mysqltmp
To configure percona-nrpe plugins for mysql you will have to do 3 things viz :
1. define commands
2. Define service
3. Add commands on client to nrpe.cfg
First 2 can be easily done for help refer http://kedar.nitty-witty.com/blog/install-percona-monitoring-tools-for-nagios-mysql-plugins#comment-43181
I found less help available for 3rd point ie to add ...
Accoing to https://www.percona.com/software/mysql-database/percona-server/faq
Percona Server is a backwards-compatible replacement for MySQL®
Default scenario is mysqldump --master-data. It will cause a downtime and duration depends on your mysql server performance, tables engine and some other factors.
If you can't afford a downtime you may try pt-table-...
Disclosure: I work for Percona. I'd prefer to comment but I don't have enough points on server fault.
You might find this discussion about compatibility on the Percona blog of interest https://www.percona.com/blog/2013/10/04/the-benefits-of-a-drop-in-compatible-mysql-alternative/
and also possibly this one might have information you can use in your ...
There is a discussion of this in this blog post, as things changed with 5.7.3: https://www.percona.com/blog/2017/06/27/ssl-connections-in-mysql-5-7/
The link in that post, though, has changed as I think the documentation may have been updated. This is a better page from the core manual to view (referenced from the CREATE USER syntax page): https://dev....
For anyone wondering, Copying the entire mysql data root (in my case) /var/lib/mysql AFTER shutting down mysqld worked perfectly to migrate data to another system.
Otherwise if you need replication the cluster solutions are more appropriate and faster than DRBD
So long as the root MySQL account uses unix_authauthentication plugin (so no password required for root user to log into MySQL) then you should be fine to remove the debian-sys-maint user.
Note that as of Debian 9/Stretch by default, it no longer exists of you install MariaDB from the default repos.
If you don't have unix_authauthentication enabled for the ...
You are right. The values displayed on the graph will be a rate, i.e. drops/second. Cacti collects a counter value every X minutes (5 by default), which increases monotonically. RRD will then graph counter-type values as a rate, as you have seen.
mysql -P 3307 -uroot -p # NO! I'm connected to MySQL, - host by default is localhost. If so, a client connects over unix socket, not over TCP. Thus -P option doesn't make sense.
However if you connect like this mysql -P 3307 -uroot -p -h 127.0.0.1 it will force the client use TCP and you'll get to Percona.