Merge ~sergiodj/ubuntu/+source/apache2:bug1969629-http2-empty-response-maxrequestsperchild-BIONIC into ubuntu/+source/apache2:ubuntu/bionic-devel
Status: | Merged |
---|---|
Merge reported by: | Sergio Durigan Junior |
Merged at revision: | 787dfeff4d3c4666fd655ac0b8832be068987648 |
Proposed branch: | ~sergiodj/ubuntu/+source/apache2:bug1969629-http2-empty-response-maxrequestsperchild-BIONIC |
Merge into: | ubuntu/+source/apache2:ubuntu/bionic-devel |
Diff against target: |
92 lines (+70/-0) 3 files modified
debian/changelog (+8/-0) debian/patches/mod_http2-Don-t-send-GOAWAY-too-early-when-MaxReques.patch (+61/-0) debian/patches/series (+1/-0) |
Related bugs: |
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
Bryce Harrington (community) | Approve | ||
Canonical Server | Pending | ||
Review via email: mp+420581@code.launchpad.net |
Description of the change
This MP fixes bug #1969629.
This bug affects apache2 on Bionic, Focal and Impish. It's an issue that happens when (a) HTTP2 is being used and (b) the MaxRequestsPerChild limit is reached. In this scenario, apache2 will send a GOAWAY packet to the client too early in the connection, which will eventually lead to no data being transferred.
This bug has been reported and fixed upstream:
https:/
https:/
I backported the patch above, and in the process I had to drop a few things from it:
- I dropped all the README/changelog update bits from the patch, as usual.
- I decided to drop the version bump of mod_http2. I don't think it makes sense to adjust the module version because of this patch specifically.
- None of the apache2 packages available on B/F/I have the h2_workers_
You can find instructions on how to reproduce the bug and test the package in the SRU template.
There's a PPA with the proposed changes here:
https:/
autopkgtest is OK:
autopkgtest [04:05:36]: @@@@@@@
run-test-suite PASS
duplicate-
htcacheclean PASS
default-mods PASS
ssl-passphrase PASS
check-http2 PASS
chroot PASS
https:/
Reviewing the bug filed to launchpad and upstream, looks like a straightforward problem with a well-defined solution, which makes sense to SRU.
I set up the three LXC containers and ran the test cases on all three releases, verified the error prints "BUG DETECTED", then installed the PPA and verified test case runs without exiting, as documented for the expected behavior.
Code review of the patch looks ok, and the changes dropped compared with the upstream version of the patch makes sense.
SRU text is quite well written; the test case is super paint-by-numbers and I really like how you suggested courses of action in case of regression in the where problems could occur section.
Nothing at all to complain about. LVGTM, +1.