great tutorial, one question: for multi master achitecture, each member provide their own endpoint to application, group replication does not provide single endpoint and route inside itsself? so in early design, developers has to decide which member to use for application or they have to use router, HAProxy to do it themselves?
Hi Chen. In this video, 👇 you can find information what could solve your question. ruclips.net/video/Mzks0isSqMM/видео.html Also, do not forget to subscribe to our channel, every week we upload new videos and database tutorials😀 ⏩ percona.tv/subscribe
Hi Des. Multi-master in PXC and group replication is not a recommended approach. In both solutions using synchronous replication is recommended to have a single Primary, with the others serving reads and ready to become primary in case of failover. This choice has its reason in years of experience, that indicate that using multi-master is not improving performance, while increasing the need to perform data certification and conflict resolution. Keep in mind an important difference between PXC and group replication. In PXC all nodes are primary by default, but we use only one node at a time to adhere to the best practices. In GR you can have two setups, single Primary or multi-primary, the default is single Primary for the already explained reason. So... A solution using synchronous replication imposes all the nodes to be aligned, data-wise, while a solution using asynchronous replication allows having the node in a different data state. The first IS NOT superior solution respect the second, is a completely different solution. The use of one or the other is guided by the need for high availability and performance. Like👍 Comment💬 Share🔗 Subscribe✅ and tap the 🔔 to be the first to see new videos on #OpenSource #Databases & more! percona.tv/subscribe
great tutorial, one question: for multi master achitecture, each member provide their own endpoint to application, group replication does not provide single endpoint and route inside itsself? so in early design, developers has to decide which member to use for application or they have to use router, HAProxy to do it themselves?
Hi Chen. In this video, 👇 you can find information what could solve your question. ruclips.net/video/Mzks0isSqMM/видео.html
Also, do not forget to subscribe to our channel, every week we upload new videos and database tutorials😀 ⏩ percona.tv/subscribe
With the multi-master setup, can we say that this is more superior that the traditional master-slave and there will be less use case for that?
Hi Des.
Multi-master in PXC and group replication is not a recommended approach. In both solutions using synchronous replication is recommended to have a single Primary, with the others serving reads and ready to become primary in case of failover.
This choice has its reason in years of experience, that indicate that using multi-master is not improving performance, while increasing the need to perform data certification and conflict resolution.
Keep in mind an important difference between PXC and group replication.
In PXC all nodes are primary by default, but we use only one node at a time to adhere to the best practices.
In GR you can have two setups, single Primary or multi-primary, the default is single Primary for the already explained reason.
So... A solution using synchronous replication imposes all the nodes to be aligned, data-wise, while a solution using asynchronous replication allows having the node in a different data state.
The first IS NOT superior solution respect the second, is a completely different solution. The use of one or the other is guided by the need for high availability and performance.
Like👍 Comment💬 Share🔗 Subscribe✅ and tap the 🔔 to be the first to see new videos on #OpenSource #Databases & more! percona.tv/subscribe