RFE - Sync object grouping and intervals

[brainfunked]

Node agent (as of 1.5.2) deals with a few different types of data sync cycles:

• Hardware related inventory, including disks, block devices, network interfaces etc.
• Service detection for known components, including tendrl components.
• Storage system detection, for when the node isn't part of an imported cluster.
• Point in time stats for CPU and memory usage.

All of the above currently happen through the same sync cycle. It may be possible to group the functionalities into different sync cycles as an optimisation:

• On startup: service detection, storage system detection, hardware inventory
• Every 10 seconds: point in time utilisation stats, service status check (sync only when changed)
• Every 3 to 5 minutes: storage system detection, hardware inventory

The storage system detection can be turned off when the cluster the node is a part of is imported.

[julienlim]

Looking to optimize the backend performance issues.

[r0h4n]

the issue here tries to make a distinction between the sync intervals of objects which are of higher interest (disks) to the Tendrl user vs the lesser important ones (network etc). This is the polishing work required for getting to a higher maturity in our data in terms of capturing what really matters to the storage admin and minimizing noise

@julienlim could you give more context about your comment? Seems out of place

[julienlim]

@r0h4n Per @nthomas-redhat In a recent status meeting/discussion, @nthomas-redhat pointed to this bug as the one related to backend performance issues so I was noting that down here per what @nthomas-redhat said as I was trying to locate the bug(s) related to the backend performance issues.