Im currently evaluating our future MySQL Architecture. We already had an Percona Consultant here, so we decided to stick with asynchronous write replication.
Therefore we are currently looking to introduce MHA or Orchestrator. Because MHA is itself not trivial runnable in HA mode and because of:
I’m highly interested in Orchestrator at the Moment, but also the setup is more complicated.
Now I also took a look at PMM and it is bundled with an Orchestrator version. Is the bundled Orchestrator meant as a fully blown Replication Manager which also handles failovers?
If yes, is it possible to “cluster” two instances of PMM together somehow?
Regarding clustering of PMM, at this time we do not support any interaction between multiple PMM server instances. Since your post focuses on Orchestrator, in what use case would you like to see Orchestrator “clustered”? for HA? If that is the concern, have you reviewed: [url]https://github.com/github/orchestrator/blob/master/docs/high-availability.md[/url]
Which parts of this HA plan would you like to see supported by PMM?
A little over a year after the original post, I would like to see if there is still any interest or if any work concerning this has been done.
I personally would like to see HA for Orchestrator via a shared backend because that seems pretty straightforward to me, though I do not know how/if multiple orchestrator instances could work on a single PMM Docker instance.
The HA raft option seems interesting too.
I would also be interested in what efforts can be made to make PMM itself (docker container especially) highly available.
These are great points! Percona is also interested in making PMM Server highly available - what approaches do you think we ought to consider for PMM docker container?
Specifically for Orchestrator we are still shipping with PMM Server but we are disabling by default, since our attention is primarily on the Monitoring component of PMM, and this isn’t forecasted to change over the coming months.