Improve Total Blocking Time performance metric of Merge Request Changes page with large amount of changes
Important
- This issue is meant to track the impact of the solutions put in place. Will be closed when the TBT metric is within acceptable ranges
- For a concrete solution we're attempting, see issue #321691 (closed) please
Proposal
Before we shipped the fix to #241841 (closed), this was the result of the MR Changes tab for 10k architecture reference:
| NAME | FCP (ms) | LCP (ms) | TBT (ms) | SI (ms) | LVC (ms) | TFR SIZE (kb) | SCORE | RESULT |
|---|---|---|---|---|---|---|---|---|
| `web_project_merge_request_changes | 1538 | ✘ 12542 (<12500) | ✓ 26156 (<32500) | 9963 | 12616 | 1352.1 | 73 | FAILED |
After this is how it looked.
| NAME | FCP (ms) | LCP (ms) | TBT (ms) | SI (ms) | LVC (ms) | TFR SIZE (kb) | SCORE | RESULT |
|---|---|---|---|---|---|---|---|---|
| web_project_merge_request_changes | 1826 | ✓ 9898 (<17500) | ✓ 26222 (<32500) | 8301 | 16117 | 1179.9 | 71 | Passed |
Target metric to improve
Taking a closer look at the TBT, it still fluctuates a look soaring to 30s frequently.
We need to bring down the TBT value.
More details
See this thread for context: #241841 (comment 469085171)
Edited by André Luís