[Koha-bugs] [Bug 26854] New: stage-marc-import.pl does not properly fork

[bugzilla-daemon at bugs.koha-community.org]

https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=26854

            Bug ID: 26854
           Summary: stage-marc-import.pl does not properly fork
 Change sponsored?: ---
           Product: Koha
           Version: master
          Hardware: All
                OS: All
            Status: NEW
          Severity: major
          Priority: P5 - low
         Component: MARC Bibliographic record staging/import
          Assignee: koha-bugs at lists.koha-community.org
          Reporter: dcook at prosentient.com.au
        QA Contact: testopia at bugs.koha-community.org

When stage-marc-import.pl forks, the child process closes STDOUT, but it does
not close STDERR, which is also connected to the Apache httpd parent process.

This means that the parent stage-marc-import.pl process becomes a zombie
process, as Apache httpd (apache2 2.4.25 Debian 9 && apache2 2.4.29 Ubuntu
18.04 && apache2 2.4.33 openSUSE Leap 15.1) does not reap it, as you can see
below:

kohadev+ 32716 32712 74 02:17 ?        00:00:26 [stage-marc-impo] <defunct>
kohadev+ 32725     1 28 02:18 ?        00:00:02 /usr/bin/perl
/kohadevbox/koha/tools/stage-marc-import.pl

The above happens when running CGI-only Koha or Plack-enabled Koha which does
not proxy stage-marc-import.pl.

Across 3 different Koha instances, I am experiencing different outcomes in the
web browser, although ultimately the forked child process does stage the
records (unless bug 26853 occurs).

--

That said, I have noticed something interesting. On openSUSE with CGI, I'm
seeing a 200 in the browser, although the AJAX request does not complete, so
the polling of background-job-progress.pl does not happen. On Ubuntu 18.04 with
Plack, Apache returns a 502 to the browser. On Debian 9 with Plack
(koha-testing-docker), no status is returned but it doesn't timeout. 

Possibly differences include different Apache httpd behaviour and different
Apache configuration. If we fix this issue, I think that we'll find that
behaviour across the different instances becomes more consistent.

-- 
You are receiving this mail because:
You are watching all bug changes.
You are the assignee for the bug.