Closed philbooth closed 5 years ago
shane-tomlinson
commented
5 years ago The status endpoint gets hit quite heavily and it's questionable how useful the perf data we are getting from it is. We could drastically reduce our daily flow event count if we stopped emitting the perf event on that route.
Since this endpoint is hit so often, could overall perf results be skewed? Even if no, removing perf results from here seems reasonable to cut down on overall flow event count.
philbooth
commented
5 years ago could overall perf results be skewed?
There's no "overall" perf result in that sense, unless someone's written a query to aggregate them that I don't know about. They're just plotted on a chart per-endpoint.
philbooth
commented
5 years ago Just to quantify the effects of this before I open a PR:
SELECT COUNT(*), SUBSTRING(type, 19) AS route
FROM flow_events
WHERE type LIKE 'route.performance.%'
GROUP BY 2
ORDER BY 1 DESC;| count | route |
|---|---|
| 530,672,699 | /recovery_email/status |
| 26,197,101 | /password/forgot/status |
| 10,845,530 | /account/device |
| 7,576,774 | /account/keys |
| 6,984,281 | /account/login |
| 6,974,785 | /recovery_email/verify_code |
| 4,377,133 | /sms/status |
| 4,144,436 | /account/create |
| 1,250,179 | /sms |
| 1,160,813 | /recovery_email/resend_code |
| 669,600 | /password/forgot/send_code |
| 619,757 | /recoveryKey/exists |
| 486,950 | /password/forgot/verify_code |
| 485,673 | /account/reset |
| 291,369 | /recovery_emails |
| 275,596 | /totp/exists |
| 238,183 | /session/destroy |
| 215,298 | /totp/create |
| 141,497 | /signinCodes/consume |
| 120,245 | /session/verify/totp |
| 117,023 | /session/reauth |
| 68,981 | /account/login/send_unblock_code |
| 50,950 | /account/device/destroy |
| 35,663 | /session/status |
| 35,192 | /password/forgot/resend_code |
| 14,765 | /recovery_email |
| 6,502 | /recoveryKey |
| 6,075 | /account/destroy |
| 4,674 | /session/verify/recoveryCode |
| 4,261 | /recovery_email/destroy |
| 1,610 | /account/device/commands |
| 1,313 | /recovery_email/set_primary |
| 779 | /session/verify/token |
| 700 | /totp/destroy |
| 558 | /recoveryCodes |
| 375 | /account/devices/notify |
| 3 | /recoveryKey/c21e1bd35a224cc39a26c98c62e2d170 |
So that's a lot of events we'll get rid of.
(and there's a long tail of about 15k different variations of that last recovery key endpoint, which I'll also clean up for the sake of tidiness)
The status endpoint gets hit quite heavily and it's questionable how useful the perf data we are getting from it is. We could drastically reduce our daily flow event count if we stopped emitting the perf event on that route.