Offline Client Possibilities

Description:

In a networking environment with a lot of segmentation, the client server approach can be problematic. Either the option is to run many PMM Servers (which is a maintenance issue), or to have the PMM Server connect to the databases.

I do see that work has been done to make PMM resilient to network faults, however, this is specifically in reference to a situation where the PMM server will never have network access to the database.

Is it possible to have a completely offline client? I do understand this would cause delays, but, would allow using PMM in the previously described environment.

In my mind, what it would look like is having a client near the database handling the job of monitoring. It would generate regular files, which could either be uploaded to somewhere common (S3), or collected periodically and loaded into PMM Server.

Again, I do completely understand this breaks many use cases (it’s slow, so alerting wouldn’t work), but would enable usage of PMM.

Also, apologies if this is covered somewhere that I’ve overlooked.

Thanks!

Hello @Brad_Misterek
Thank you for the details on your use case for PMM and requirements.

The functionality you described is not available in PMM, But this could be a Feature Request, and you can report it here: https://perconadev.atlassian.net

After creating FR, Our engineering team will review the requested use case of this feature Request.

Hello @Brad_Misterek,
PMM Client can collect metrics offline during long network faults and send them to PMM Server once connection is available.
I’m not sure if we are going to implement completely offline client at all. We are working on other HA options. Here is blogpost written by our PM High Availability: Choosing the Right Option for Your Percona Monitoring and Management. I hope it will help you.

1 Like