PMM server High Availability

Do we have any case study for how we can achieve High Availablity for our PMM server. In case we lose the VM on which PMM is present, we would be losing a lot of Metrics data. Metrics data and alert system will be all at halt! Do we have any suggestions on how can we achieve High Availability for this PMM server with a minimal time frame of metrics data lost?

Hi gkmgaurav , this is a great question! Percona hasn’t published a case study on delivering High Availability with PMM. In general the recommended way to protect against data loss is to take a backup of the PMM data container with:

sudo docker save -o /backup/pmm-data-`date +%Y%m%d`.tar pmm-data

This will at least give you a restore point should disaster strike PMM Server.

We are actively considering other options with regards to High Availability, and as we solidify our path we will blog about it! Thanks for the question.

Thanks Michael. Taking backup at regular intervals is always necessary. But, if PMM can have HA, that would be great!!

I believe this is theoretically possible, because Prometheus has HA. And Grafana is just a client. But this can only be done by Percona…