Succes Story Argelith
The high availability solution with MySQL for CANIAS ERP
About acar software & FromDual
The acar Software GmbH cooperates for several years with the FromDual GmbH in the field of high available solutions of databases. They guarantee the customer with MySQL a high available solution for CANIAS ERP, whereby the acar software GmbH takes the part of CANIAS and FromDual takes the part of MySQL. As a team they design the high availability for a failsafe enterprise application.
The Argelith Bodenkeramik GmbH become a global Player around the industrial tile manufacturer because of their locations in Germany, USA and Russia. The daily work in different time zones requires a failsafe IT operation. During the ERP induction the advantages of the ERP Sytem CANIAS were realized and were fast implemented into the optimization of the business process. Because of the failsafety it was intended to guarantee that these processes were not endangered because of unnecessary downtimes or maintenance window. Furthermore it was the goal to install the IT landscape in a way without any administration effort to guarantee that the ERP operation could be presented, because there should not be build a 24 hour IT-Support.
The MySQL based solution
In order to guarantee the failure safety a activ/passiv failover cluster was chosen after a locally analysis. The cluster was realized at two own local divided datacenters in order to guarantee the failure safety against environmental influences like fire and water. It was an additional challenge for the team to work in different time zones (Europe, USA and Russia). Timeframes for maintanance, backup and the high availability are loom large.
The technical concept behind
The concept behind a activ/passiv failover-cluster is as followed: You take 2 Server (also named node) which communicates via a so called cluster software. DRBD synchronize the data synchronic from Server 1 to Server 2. The MySQL Database runs only on one node (active). Durring this time the second node (passiv) waits and accepts only the data via DRBD. The Database on the active node is called via a virtual IP (VIP). If an error occurs this error is noticed by the cluster software and a failover of the resource (VIP inclusive) is lead to the passive node. There the database is available after a automatically recovery after a few minutes. The whole failover is completely transparent for the CANIAS ERP. Theoretically every application can be make high available with this kind of setup. Such a active/passiv failover-cluster (also called MySQL HA, not to mix up with MySQL Cluster!) is associated with a availability of 99.99% (4×9) HA normally. This makes a downtime of 52 minutes per year.
The following instruction based on MySQL 5.1 (the database version is not really important in this case). The used operating system is Linux/CentOS 5.5, but other Linux distributions works as well. DRBD v8.3.8 and Heartbeat v3.0.3 was used additionally.