32k truncation issue on master branch

[ludoch]

See https://code.google.com/p/googleappengine/issues/detail?id=12790&q=32k&colspec=ID%20Type%20Component%20Status%20Stars%20Summary%20Language%20Priority%20Owner%20Log
Issue is sovled in async branch, but need resolution on master as more customers are using it now, and it's getting escalated.
@gregw : can you take an urgent look? It seems it might be possible to address according the bug above?

[joakime]

Looking into this.

[joakime]

The reproduction examples provided in that code.google.com link to reproduce are not reproducible directly (outside of Spring / Groovy / Google)

Test case added to branch bugs/316 uses the CommitDelayResponse object in an effort to trigger this bug.

It does not trigger with sizes:

size / notes
32768 / 32k on the nose
32668 / 32k - 100
32767 / 32k - 1
32868 / 32k + 100
32769 / 32k + 1

Do we know what the CommitDelayResponse configuration is when the issue occurs?
Does this occur with standard HTTP or HTTPS? or both?

Will endeavor to find the cause.

Incidentally, the CommitDelayResponse implementation (in master) has a spec breaking implementation of sendError() and sendRedirect() (unrelated to this specific bug)

[ludoch]

Thanks for looking into this one... Do you have permission to post directly
on the original bug? If not, I'll see if I can add you

On Mon, Aug 8, 2016 at 12:28 PM, Joakim Erdfelt [email protected]
wrote:

The reproduction examples provided in that code.google.com link to
reproduce are not reproducible directly (outside of Spring / Groovy /
Google)

Test case added to branch bugs/316 uses the CommitDelayResponse object in
an effort to trigger this bug.

It does not trigger with sizes:

size / notes
32768 / 32k on the nose
32668 / 32k - 100
32767 / 32k - 1
32868 / 32k + 100
32769 / 32k + 1

Do we know what the CommitDelayResponse configuration is when the issue
occurs?
Does this occur with standard HTTP or HTTPS? or both?

Will endeavor to find the cause.

Incidentally, the CommitDelayResponse implementation (in master) has a
spec breaking implementation of sendError() and sendRedirect() (unrelated
to this specific bug)

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#316 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAE4zUtD12uPtmfJ_lL9GvfIFptgUZoxks5qd4NCgaJpZM4JfOzP
.

[gregw]

Ludo,

Do you want us to look into removing the 32KB limit from master? This
would be doable, but would basically require redoing most of what is in the
async-support branch.

Remind me again what is the main issue why we cannot make async-support the
master branch?

cheers

On 9 August 2016 at 09:37, Ludovic Champenois [email protected]
wrote:

Thanks for looking into this one... Do you have permission to post directly
on the original bug? If not, I'll see if I can add you

On Mon, Aug 8, 2016 at 12:28 PM, Joakim Erdfelt [email protected]
wrote:

The reproduction examples provided in that code.google.com link to
reproduce are not reproducible directly (outside of Spring / Groovy /
Google)

Test case added to branch bugs/316 uses the CommitDelayResponse object in
an effort to trigger this bug.

It does not trigger with sizes:

size / notes
32768 / 32k on the nose
32668 / 32k - 100
32767 / 32k - 1
32868 / 32k + 100
32769 / 32k + 1

Do we know what the CommitDelayResponse configuration is when the issue
occurs?
Does this occur with standard HTTP or HTTPS? or both?

Will endeavor to find the cause.

Incidentally, the CommitDelayResponse implementation (in master) has a
spec breaking implementation of sendError() and sendRedirect() (unrelated
to this specific bug)

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<https://github.com/GoogleCloudPlatform/appengine-
java-vm-runtime/issues/316#issuecomment-238348922>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAE4zUtD12uPtmfJ_
lL9GvfIFptgUZoxks5qd4NCgaJpZM4JfOzP>
.

—
You are receiving this because you were assigned.
Reply to this email directly, view it on GitHub
#316 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAEUrabBzKMhAYbkBoO7wdAFwkaRG41mks5qd73VgaJpZM4JfOzP
.

Greg Wilkins [email protected] CTO http://webtide.com

[ludoch]

Main issues on async branch:

1/ request logs: The problem is that the runtime logs are not available in request_log like they normally do. Data is still available but as far as I checked those reside in app logs.
https://code.google.com/p/googleappengine/issues/detail?id=13170

2/ Some new API exceptions https://code.google.com/p/googleappengine/issues/detail?id=13169 (might or might not be Jetty, but need further investigation)

3/ Crashes https://code.google.com/p/googleappengine/issues/detail?id=13168 Again, might not be related to async work, but blocking some customers.

Main pain point is number one. But I guess it is inherent to the async work? Can it be improved?

[gregw]

For 1/ I have created #318 , note also that we have #231 outstanding on the logging as well, which is waiting for a bit of testing. Will look into both shortly.

[gregw]

closing as this issue became rather meta and issues are discussed elsewhere.
@ludoch please reopen if there are changes you want for logging in master branch of this repo.