Re: Apache hangs after having 6 x SSE streams active

Mod_perl won't work with mpm_event and threads unless you compile Perl with s=
pecific options and avoid doing certain things which can cause segmentation f=
aults in Perl. You can't just use an "off-the-shelf" version of Perl. Search=
 the mod_perl mailing list archives for posts by Joe Schaefer going back ove=
r the past ~3 years for clues on how to do that and check out his sealed.pm m=
odule. I haven't done it or used it personally. It sounds non-trivial to set=
 up properly. If you try going down this route, please report back with your=
 findings.

According to conventional wisdom, mpm_prefork works better (or at least easi=
er) for mod_perl and is often recommended for mod_perl. Most mod_perl develo=
pers avoid using threads entirely because the combination of mod_perl and th=
reads is so problematic. Mpm_prefork is less efficient than mpm_event by mos=
t accounts, of course. You would certainly need to configure Apache to have w=
ay more than 4 servers with mpm_prefork.

Good luck,
Ed

> On Oct 23, 2025, at 11:42=E2=80=AFAM, Steven Haigh via modperl <modperl@pe=
rl.apache.org> wrote:
> I'm writing a web UI in perl + mod_perl that sends an SSE stream to the br=
owser full of JSON formatted updates etc. The browser then renders that all v=
ia Javascript to the user.
>=20
> I've noticed that if I reload the page multiple times, or just navigate aw=
ay, when there's ~6 existing SSE streams, apache will stop accepting new con=
nections completely.
>=20
> I've hunted around via the normal search engines, turned up very little - a=
nd out of desperation, asked ChatGPT's Codex...
>=20
> It suggested the following apache config:
>=20
> LoadModule mpm_event_module modules/mod_mpm_event.so
> # Tune event MPM so we have headroom for long-lived SSE connections
> <IfModule mpm_event_module>
>    ServerLimit              4
>    StartServers             2
>    MinSpareThreads         50
>    MaxSpareThreads        150
>    ThreadsPerChild         64
>    MaxRequestWorkers      256
>    MaxConnectionsPerChild   0
> </IfModule>
>=20
> ## Load mod_perl
> LoadModule perl_module /usr/lib/apache2/modules/mod_perl.so
> PerlModule Apache2::Reload
> PerlInitHandler Apache2::Reload
> PerlOptions +Parent
> PerlSwitches -Mlib=3D/usr/local/apache2/htdocs-includes
> PerlInterpStart      20
> PerlInterpMinSpare   10
> PerlInterpMaxSpare   20
> PerlInterpMax        80
>=20
> Even with this however, when I hit 6 x concurrent SSE streams being served=
, Apache grinds to a halt and doesn't even load the /server-status/ page for=
 troubleshooting.
>=20
> Once some of the 'dead' SSE streams fail and those connections terminated,=
 apache starts behaving again.
>=20
> Has anyone come across this before and maybe knows a bit of additional kno=
wledge as to why this is, and what I can do about it?