The perf regression sheriff tracks performance regressions in Chrome's continuous integration tests. Note that a new rotation has been created to ensure the builds and tests stay green, so the perf regression sheriff role is now entirely focused on performance.
Open the perf dashboard alerts page.
In the upper right corner, sign in with your Chromium account. Signing in is important in order to be able to kick off bisect jobs, and see data from internal waterfalls.
Pick up Chromium Perf Sheriff from “Select an item ▼” drop down menu. table of “Performance Alerts” should be shown. If there are no currently pending alerts, then the table won't be shown.
The list can be sorted by clicking on the column header. When you click on the checkbox next to an alert, all the other alerts that occurred in the same revision range will be highlighted.
Check the boxes next to the alerts you want to take a look at, and click the “Graph” button. You'll be taken to a page with a table at the top listing all the alerts that have an overlapping revision range with the one you chose, and below it the dashboard shows graphs of all the alerts checked in that table.
During your shift, you should try to follow up on each of the bugs you filed. Once you've triaged all the alerts, check to see if the bisects have come back, or if they failed. If the results came back, and a culprit was found, follow up with the CL author. If the bisects failed to update the bug with results, please file a bug on it (see feedback links below).
Also during your shift, please spend any spare time driving down bugs from the regression backlog. Treat these bugs as you would your own -- investigate the regressions, find out what the next step should be, and then move the bug along. Some possible next steps and questions to answer are:
When a bug does need to be pinged, rather than adding a generic “ping”, it's much much more effective to include the username and action item.
You should aim to end your shift with an empty backlog, but it's important to still advance each bug in a meaningful way.
After your shift, please try to follow up on the bugs you filed weekly. Kick off new bisects if the previous ones failed, and if the bisect picks a likely culprit follow up to ensure the CL author addresses the problem. If you are certain that a specific CL caused a performance regression, and the author does not have an immediate plan to address the problem, please revert the CL.
Perf regression sheriffs have their eyes on the perf dashboard and bisects more than anyone else, and their feedback is invaluable for making sure these tools are accurate and improving them. Please file bugs and feature requests as you see them:
Speed>Bisection. If a bisect problem is blocking a perf regression bug triage, please file a new bug with component
Speed>Bisectionand block the regression bug on the bisect bug. This makes it much easier for the team to triage, dupe, and close bugs on the infrastructure without affecting the state of the perf regression bugs.
Speed>Benchmarksand cc the owner.