Bug 60647 - "access this connection" wrong with event MPM
Summary: "access this connection" wrong with event MPM
Status: RESOLVED FIXED
Alias: None
Product: Apache httpd-2
Classification: Unclassified
Component: Core (show other bugs)
Version: 2.5-HEAD
Hardware: PC Linux
: P2 normal (vote)
Target Milestone: ---
Assignee: Apache HTTPD Bugs Mailing List
URL:
Keywords: FixedInTrunk
Depends on:
Blocks:
 
Reported: 2017-01-25 20:30 UTC by Eric Covener
Modified: 2019-05-27 23:21 UTC (History)
1 user (show)


Attachments
keep conn_count more accurate (2.43 KB, patch)
2017-01-25 20:30 UTC, Eric Covener 		Details | Diff
remove test val (2.39 KB, patch)
2017-01-25 21:56 UTC, Eric Covener 		Details | Diff
Show Obsolete (1) View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Eric Covener 2017-01-25 20:30:43 UTC 
Created attachment 34676 [details]
keep conn_count more accurate

The first field of ACC in /server-status is requests on the current connection.

With event, a connection hops from slot to slot, only visible while it's being processed.

These values are currently always zero w/ trunk and 2.4 under event.

ap_increment_counts() tracks this stuff per-slot and event causes it to be blown away in worker_thread() right before process_socket.

I have a patch that just allows ap_process_http_async_connection to push a new "requests on current connection" onto the scoreboard handle for the slot its currently using.  This is a bit of an odd place to fix up the connection count, but the ap_increment_counts is currently only called by the eor_bucket and I didn't want to put async vs. not async logic there.
Comment 1 Eric Covener 2017-01-25 21:56:20 UTC 
Created attachment 34678 [details]
remove test val
Comment 2 Graham Leggett 2019-05-27 23:21:37 UTC 
Backported to v2.4.40.