Feature #55278: mgr/telemetry: implement caching when generating the telemetry report - mgr - Ceph

Actions

Copy link

Feature #55278

open

mgr/telemetry: implement caching when generating the telemetry report

Added by Laura Flores about 2 years ago. Updated about 2 years ago.

Status:

New

Priority:

Normal

Assignee:

Category:

telemetry module

Target version:

% Done:

Source:

Tags:

Backport:

Reviewed:

Affected Versions:

Pull request ID:

Description

On large clusters, it can take up to 3 minutes to generate a Telemetry report. There are several ways we can combat this:

1. Cache, or pre-gather, the Telemetry report ahead of time

This can be done instead of generating the report from scratch each time a `show` or `preview` command is triggered. As a result, users may see a slightly outdated version of the report; however, the ultimate goal of the `show` and `preview` is to allow users to see which metrics they are reporting or can report. Slightly outdated metrics is negligible if the report can be generated faster.

Relevant commands:
- `ceph telemetry show/preview`
- `ceph telemetry show/preview-device`
- `ceph telemetry show/preview-all`
- `ceph telemetry show/preview <channel_name>`

2. Implement a "summary" CLI command to generate a shorter version of the report

This solution could be done in addition to caching. Some users may only be interested in viewing a summarized version of the report instead of every collected metric. The command could be something like `ceph telemetry show/preview summary`. See `ceph pg dump summary` for inspiration.