This blog post is a continuation of the
infrastructure load series that John O'Duinn had been writing since 2009.
For now, I will be doing this every month and I will be providing a new format to help me see the big picture more easily. If there's anything you're missing from the posts that John used to do please let me know.
Feedback very much wanted! :)
Link to the 2013-12 pushes report.
Highlights
- Pushes/month: 6,063 pushes
- This is a low number compared to our usual load
- This is due to developers being off since Dec. 20th
- We've had over 7000 pushes/month in the last 6 months
- except October with the summit
- Even then, December had more pushes than February and June of 2013
- Pushes/day:
- 196 pushes/day
- 14 days over 300 pushes
- Pushes/hour:
- On average: 9 hours of the day had more than 9 pushes/hour
- On average: our busiest hours were between 8am and 3pm
- On average: we had ~12 pushes between 13:00-14:00 Pacific time
- The top 4 branches were:
- try (50%)
- mozilla-inbound (17.4%)
- b2g-inbound (10.2)
- fx-team (6.8%)
- They accounted for 84.4% of all pushes
Observations
- Since the beginning of the year we have seen a decline on number of pushes for mozilla-inbound
- This is due that the fx-team and the b2g-inbound trees are sheriff managed trees
- The new pool of Windows machines were added around May of 2013
- Two months later we see a substantial growth on the Try tree usage
- I assume that this helped testing on the Try tree
- We were under-water-level with our old Windows test pool
- NOTE: Testing on the Try tree has the lowest priority when scheduling jobs
- I hope this means that developers saw the Try tree to be useful again
Current records
- August of 2013 was the month with most pushes (7,771 pushes)
- August 26th had the highest number of pushes in one day with 443 pushes
- Our highest average of "pushes-per-hour" is 15.73 pushes/hour
This work by
Zambrano Gasparnian, Armen is licensed under a
Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
