Hot answers tagged

16

Because like any other optimization, it doesn't fit every workload. Galera can get overwhelmed by a high rate of transactions, or when transactions update many rows. It can also make your applications experience delays on COMMIT as the cluster is synced up. Galera doesn't update other nodes synchronously, either. It just transmits worksets synchronously. ...


10

The problem is actually very simple. Here is what happened When you installed MySQL, the 5 InnoDB systems tables exist in two places inside /var/lib/mysql/mysql as 5 .frm and 5 .ibd files inside the data dictionary within ibdata1 (InnoDB System Tablespace) At some point in your installation, you must have deleted ibdata1. This left the 10 InnoDB system ...


7

MariaDB Galera Cluster: Solution 1: 1) I've changed safe_to_bootstrap parameter to 1 on one of the node in the file /var/lib/mysql/grastate.dat: safe_to_bootstrap: 1 2) After that I killed all mysql processes: killall -KILL mysql mysqld_safe mysqld mysql-systemd 3) And started a new cluster: galera_new_cluster 4) All other nodes I reconnected to the ...


6

If you want real HA ("High Availability"), you need 3 nodes in 3 separate locations. When one machine goes down, the rest of the nodes vote and see that they are in the majority. They see 2 out of 3, so they declare themselves to be the cluster and their sibling to be out of the cluster. If the dead machine comes back online, the majority chatter with the ...


6

There was a bug in galera_recovery.sh script. https://jira.mariadb.org/browse/MDEV-10396


6

but I have difficulties to see the added value compared to Cloud SQL instances with replication and backup. I am not very familiar in scaling heavy load systems. Galera-based clusters like Percona XtraDB Cluster support true active/active multi-master, so failover is seamless because you can actually be writing on any node at any time (tho it's ...


5

1) Is a two node configuration safe\supported? Galera will still run in a two node setup. However, there is always the threat of a split-brain scenario. For example, suppose you have DB1 and DB2 form a two node Galera Cluster. If DB1 goes down, you need to failover to DB2. While DB1 is having maintenance done to bring it back up, DB2 gets all your changes. ...


4

Galera (for MySQL, Percona XtraDB Cluster or MariaDB Cluster, they are essentially the same but from different vendors and base mysql version) can work perfectly with 2 machines, and it is a very common setup for substituting standard MySQL replication. Requiring 3 nodes is not a requirement of Galera, but of any cluster that cares about data consistency ...


4

The variable you're looking for is innodb_status_output. Setting it to OFF should stop the innodb status from being dumped into the error log.


4

The error message is misleading. Galera can mark user threads as high priority which means they can't be killed. Currently, this gives the ERROR 1095 (HY000): You are not owner of thread... See MDEV-12008 for details. This will probably affect other distributions of Galera as well, such as Percona XtraDB Cluster.


3

Yes, it is very possible. As the Documentation says, you should have a minimum of three nodes in a Percona XtraDB Cluster. You only need one node out of the three nodes. Make sure all nodes in the PXC have unique server_ids. Make sure the Master you are replicating from has a unique server_id from the PXC nodes Make sure the master and every PXC node is ...


3

I have been asking the same question for several months and have also been unable to find a clear answer on the internet. Perhaps that means that I'm more concerned than I should be; perhaps I should not be doing database server maintenance... So I was forced to do this yesterday, and everything worked well. This is how I went about it: I upgraded the ...


3

You should use some kind of load-balancing solution, at least for these main reasons: To equally balance the load between the nodes. To check availability of the nodes and make it transparent as possible to your app. If you use an HAproxy server you should have it replicated, so it's not a single point of failure, e.g. you can have two HAproxies in active/...


3

In order of your topics: You can't do that with cluster replication. Multi-master is a cluster feature. You should try to use the replicate_wild_do_table setting. Slave threads will be restricted to replicating tables that match the specified wildcard pattern. For example replicate-wild-do-table=foo%.bar% will replicate only updates to tables in all ...


3

If wsrep_local_state_comment, wsrep_ready, and wsrep_connected show the correct values, then that particular node is good. There's no Seconds_Behind_Master equivalent, because Galera is synchronous -- a healthy node can't be lagging, by definition and by design... a slow node slows down all the others, it doesn't lag behind them.


3

At MariaDB 10.1 where Galera included you need to start first node with key --wsrep-new-cluster e.g. /etc/init.d/mysql start --wsrep-new-cluster All other nodes starts as usually: /etc/init.d/mysql start To determine which node needs to be bootstrapped, compare the wsrep_last_committed value on all DB nodes: KVM-1> SHOW STATUS LIKE 'wsrep_%'; +-------...


3

Pre-Recovery Settings: Ensure that the MYSQL_HOME path is being exported in the .profile. If the MySQL install is in a different location, then make that change to the MYSQL_HOME.(Example: MYSQL_HOME=/path/to/mysql) Crash Recovery Steps: Find a valid seqno. Look at the grastate.dat file on each server to see which machine has the most current data. The ...


3

I've done some research on this now, so I'll answer on my own question. Summary first If the performance problems are observed only on one node, it's perfectly safe to turn off innodb_doublewrite and innodb_flush_log_at_trx_commit on the affected node - just remember to use one of the other nodes for bootstrapping should the whole cluster fail. # echo "[...


3

wsrep_sst_donor=2.2.2.2 wsrep_sst_donor should have the node name, not the IP address. A similar case was reported in the MariaDB Jira as MDEV-13687.


3

Your node2 needs to have log-slave-updates=1. See also SeveralNines' article on the topic which states: log-slaves-updates=1 is necessary for the writes to propagate from the slave to the other galera nodes


3

OK, 3 Galera nodes, with: Each having all disk stuff sitting in RAM disks. Enough RAM to make that possible. The nodes separated geographically so that a volcano, flood, etc, cannot take out all of them at once. High speed network. Notes: That will allow for full and automatic recovery from one server loss. The only delay during a failure is switching ...


3

The "insert_id" is local to the connection. No locking is needed to get the value. This applies in all flavors of multi-processing, clustering, engines, etc. Still, you may need "transactions" for other purposes. You have not described the code enough to say whether you need to exactly replace LOCK TABLES (etc) with BEGIN..COMMIT. Or whether some ...


2

Some drawbacks of Galera include: Storage engine support: limited to InnoDB/XtraDB (plus experimental support for MyISAM) OS support: only Linux/Unix-like Oses There are also some limitations which should be noted, but can perhaps be worked around: By default (Total Order Isolation) DDL operations block the entire cluster until they complete Every ...


2

We were able to resolve this using mysqldump instead of mydumper/myloader. Apparently there is a bug in myloader or in Galera itself which prevents replication to flow normally.


2

There is not such a thing as 100% availability. Even big providers like Google or Amazon only guarantee "three nines". The most critical systems tend to give you high availability "five nines" ("guaranteed" is different from "measured"). Some software solutions, including MySQL-related ones, can provide you something around 99,999%, but with very specific ...


2

Galera cluster auto-provisions its nodes, but that is not the case with traditional replication. You must do it manually. Look at MariaDB documentation for setting up a slave from a backup. The process is essentially the same as doing it in MySQL, but the syntax is slightly incompatible. It usually just involves mysqldump with the --master-data and --gtid ...


2

Moving to Galera would be an option. Then your application has to be aware of 'deadlock detected try restarting transaction' errors. Another possibility is to set up a read slave and configure your application to use this slave for read_only queries. If one slave isn't enough you can add more and use them with a loadbalancer of any kind. Examples are ...


2

I managed to solve this -for the most part- by changing a few parameters. apparently it looks like the database needs to warm up by opening the tables at least once. This is behavior I am familiar with since mysql 5 with plenty of tables and databases. The query still runs slow when it's the first time it runs (TABLE CACHE IS OFF). By changing the ...


2

Some general considerations, first: MySQL, and I am talking about InnoDB specifically, is optimized for OLTP loads, specially for good performance for point SELECTS and writes. That doesn't mean that you cannot do bulk inserts (I get 600 000 row insertions/second on my desktop-grade SSD), but that has an impact on the performance of other operations, mainly ...


Only top voted, non community-wiki answers of a minimum length are eligible