Remove Socket Timeout from FPM Layer #772
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Member
|
Thank you! |
mnapoli
added a commit
that referenced
this pull request
Jan 4, 2022
When Lambda times out with the PHP-FPM layer, the logs written by the PHP script are never flushed to stderr by PHP-FPM. That means they never reach CloudWatch, which makes timeouts really hard to debug. With this change, Bref waits for the FPM response until 1 second before the actual Lambda timeout (via a connection timeout on the FastCGI connection). If Bref reaches that point, it will ask PHP-FPM to gracefully restart the PHP-FPM worker, which: - flushes the logs (logs end up in CloudWatch, which is great) - restarts a clean FPM worker, without doing a full FPM restart (which may take longer) Follow up of #770, #772, #895 May address some of #862 Note: this does not change anything for the Function layer (only affects FPM). Also this does not show a full stack track of the place in the code where the timeout happens (#895 did). Still it's an improvement over the current status.
mnapoli
added a commit
that referenced
this pull request
Jan 4, 2022
When Lambda times out with the PHP-FPM layer, the logs written by the PHP script are never flushed to stderr by PHP-FPM. That means they never reach CloudWatch, which makes timeouts really hard to debug. With this change, Bref waits for the FPM response until 1 second before the actual Lambda timeout (via a connection timeout on the FastCGI connection). If Bref reaches that point, it will ask PHP-FPM to gracefully restart the PHP-FPM worker, which: - flushes the logs (logs end up in CloudWatch, which is great) - restarts a clean FPM worker, without doing a full FPM restart (which may take longer) Follow up of #770, #772, #895 May address some of #862 Note: this does not change anything for the Function layer (only affects FPM). Also this does not show a full stack track of the place in the code where the timeout happens (#895 did). Still it's an improvement over the current status.
mnapoli
added a commit
that referenced
this pull request
Jan 4, 2022
When Lambda times out with the PHP-FPM layer, the logs written by the PHP script are never flushed to stderr by PHP-FPM. That means they never reach CloudWatch, which makes timeouts really hard to debug. With this change, Bref waits for the FPM response until 1 second before the actual Lambda timeout (via a connection timeout on the FastCGI connection). If Bref reaches that point, it will ask PHP-FPM to gracefully restart the PHP-FPM worker, which: - flushes the logs (logs end up in CloudWatch, which is great) - restarts a clean FPM worker, without doing a full FPM restart (which may take longer) Follow up of #770, #772, #895 May address some of #862 Note: this does not change anything for the Function layer (only affects FPM). Also this does not show a full stack track of the place in the code where the timeout happens (#895 did). Still it's an improvement over the current status.
mnapoli
added a commit
that referenced
this pull request
Feb 14, 2023
Remove Socket Timeout from FPM Layer
mnapoli
added a commit
that referenced
this pull request
Feb 14, 2023
When Lambda times out with the PHP-FPM layer, the logs written by the PHP script are never flushed to stderr by PHP-FPM. That means they never reach CloudWatch, which makes timeouts really hard to debug. With this change, Bref waits for the FPM response until 1 second before the actual Lambda timeout (via a connection timeout on the FastCGI connection). If Bref reaches that point, it will ask PHP-FPM to gracefully restart the PHP-FPM worker, which: - flushes the logs (logs end up in CloudWatch, which is great) - restarts a clean FPM worker, without doing a full FPM restart (which may take longer) Follow up of #770, #772, #895 May address some of #862 Note: this does not change anything for the Function layer (only affects FPM). Also this does not show a full stack track of the place in the code where the timeout happens (#895 did). Still it's an improvement over the current status.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
After working on #770 and #771, @mnapoli had an amazing idea of simply getting rid of the timeout altogether. It makes total sense:
No need for custom environment variable control or weird hacks on the bootstrap process. I tested this on API Gateway as well as ALB and both work fine with FPM.