Use microseconds in addition to PID. Use the sorted file list. #4
Conversation
Needs some testing.
… in the right order
|
|
||
| sorted_file_names = sorted(pd_file_names, key=file_timestamp) | ||
| return pd_file_names | ||
| sorted_file_names = sorted(pd_file_names, key=file_timestamp, reverse=True) |
There was a problem hiding this comment.
Oops - have we been sorting this wrong?! Can you show some example file names & sort results here from a python console run?
There was a problem hiding this comment.
When I was testing without reverse it seemed to always send them in the wrong order. I'll do some more concrete tests via the Python console.
There was a problem hiding this comment.
If you enqueue a few things & then either run this code by hand or run it with print/log statements, we should be able to see how the sorting goes. :)
There was a problem hiding this comment.
Q: were the tests for seconds version of the code + sends in the same second?
There was a problem hiding this comment.
I think due to the queued files and because I wasn't using the cronjob to flush the queue, I ended up sending the files wrong in the wrong order by name but they looked right when viewing the trigger/resolve API calls. I'll fix the ordering and make the few other changes.
Using time.time()'s already available microseconds
|
Updated with the suggestions from before. It looks like we weren't using the correct variable after sorting the file list. |
| @@ -142,7 +142,7 @@ def file_timestamp(file_name): | |||
| return int(re.search('pd_(\d+)_', file_name).group(1)) | |||
|
|
|||
| sorted_file_names = sorted(pd_file_names, key=file_timestamp) | |||
| return pd_file_names | |||
| return sorted_file_names | |||
|
Looks good! |
|
(title of the PR a bit misleading now, but meh :) |
mdcollins05
changed the title
|
Fixed! |
Also, we try really hard to send the events in the right order. The latest code worked 6 out of 6 tries with 1/10th of a second between calls to pagerduty.py, which were run in the background.