
Hello, if you look at this graph you'll notice how my VPS "reacted" after mailman3 install. It's a Debian stable host, and I installed packages via apt.
I mean, isn't this a bit too much for a software that is idle, because no lists are currently in use?
RED is my VPS, the other colors are other machines.
Thanks
-- roughnecks
Mailman's content filtering has removed the following MIME parts from this message.
Content-Type: image/jpeg Name: raw (1).jpg

Il 28/05/2025 10:43, roughnecks via Mailman-users ha scritto:
Hello, if you look at this graph you'll notice how my VPS "reacted" after mailman3 install.
RED is my VPS, the other colors are other machines.
Alright, no attachments: https://postimg.cc/D4bvJhBy

On 28.05.25 17:10, roughnecks via Mailman-users wrote:
Il 28/05/2025 10:43, roughnecks via Mailman-users ha scritto:
Hello, if you look at this graph you'll notice how my VPS "reacted" after mailman3 install.
RED is my VPS, the other colors are other machines.
Alright, no attachments: https://postimg.cc/D4bvJhBy
Well, that graph is pretty pointless as it doesn't say which processes are running.
Check with "top" on the VPS. What is using the CPU?
If you have opened web and mail ports to the internet it might as well just be the standard internet background noise... But that's just a guess. "top" will tell you...
-Gerald

Il 28/05/2025 17:18, Gerald Vogt ha scritto:
if you look at this graph you'll notice how my VPS "reacted" after mailman3 install.
RED is my VPS, the other colors are other machines.
Alright, no attachments: https://postimg.cc/D4bvJhBy
Well, that graph is pretty pointless as it doesn't say which processes are running.
Maybe you didn't read my first message. That graph is *exactly* showing the difference between my VPS without and with mailman running.
Check with "top" on the VPS. What is using the CPU?
It's the manage.py qcluster whatever is called. I have shut down MM for the time being, because I'm not fully motivated to have it running, so I can't be more accurate right now.
mail and web servers are running since 3 year already on that VPS.

On 28.05.25 17:27, roughnecks via Mailman-users wrote:
Il 28/05/2025 17:18, Gerald Vogt ha scritto:
if you look at this graph you'll notice how my VPS "reacted" after mailman3 install.
RED is my VPS, the other colors are other machines.
Alright, no attachments: https://postimg.cc/D4bvJhBy
Well, that graph is pretty pointless as it doesn't say which processes are running.
Maybe you didn't read my first message. That graph is *exactly* showing the difference between my VPS without and with mailman running.
Maybe you didn't read my message? External access can cause processes to run...
Check with "top" on the VPS. What is using the CPU?
It's the manage.py qcluster whatever is called. I have shut down MM for the time being, because I'm not fully motivated to have it running, so I can't be more accurate right now.
Well, it's difficult to tell you anything if you are not able to give specifics.
In general: check the logs of the process which is running. Maybe increase the loglevel for that process to get more insight.
Check incoming traffic which may trigger it.
mail and web servers are running since 3 year already on that VPS.
And? Bots maybe scanning your server for software that wasn't there for three years and you didn't really notice. But now it's there...
It's hard to tell anything with the tiny bits of information you give...
-Gerald

On 5/28/25 08:27, roughnecks via Mailman-users wrote:
It's the manage.py qcluster whatever is called. I have shut down MM for the time being, because I'm not fully motivated to have it running, so I can't be more accurate right now.
I don't know how qcluster is configured in the Debian packages (which are out of date and possibly buggy anyway). You may be using django-q <https://django-q.readthedocs.io/en/latest/> and if so, you should migrate to django-q2 <https://django-q2.readthedocs.io/en/master/>. Both typically use memcached as a backend.
Settings that control this are typically
# Set retry > timeout.
Q_CLUSTER = {
'timeout': 300,
'retry': 360,
'save_limit': 100,
'orm': 'default',
}
and possibly a setting like 'workers': 2,
(default is the CPU count of
the system but see
<https://django-q2.readthedocs.io/en/master/configure.html#f1>).
CACHES = {
'default': {
'BACKEND': 'django.core.cache.backends.memcached.PyLibMCCache',
'LOCATION': '127.0.0.1:11211',
}
}
although the default if not set is
CACHES = {
"default": {
"BACKEND": "django.core.cache.backends.locmem.LocMemCache",
}
}
It is the 'orm': 'default'
setting that says to use Django's CACHES.
See <https://django-q2.readthedocs.io/en/master/brokers.html#django-orm>.
You can get info about what qcluster is doing with manage.py qinfo
and
monitor things with manage.py qmemory
and manage.py qmonitor
.
-- Mark Sapiro <mark@msapiro.net> The highway is for gamblers, San Francisco Bay Area, California better use your sense - B. Dylan

Il 28/05/2025 23:42, Mark Sapiro ha scritto:
On 5/28/25 08:27, roughnecks via Mailman-users wrote:
It's the manage.py qcluster whatever is called. I have shut down MM for the time being, because I'm not fully motivated to have it running, so I can't be more accurate right now.
I don't know how qcluster is configured in the Debian packages (which are out of date and possibly buggy anyway). You may be using django-q <https://django-q.readthedocs.io/en/latest/> and if so, you should migrate to django-q2 <https://django-q2.readthedocs.io/en/master/>. Both typically use memcached as a backend.
I wouldn't know how to migrate and if that means doing anything not in the debian-way (haven't opened any link so far).
About qcluster configs, yes, all correct but there is no configuration for caches, so probably it's the default one you specified. I can look into this.
You can get info about what qcluster is doing with
manage.py qinfo
and monitor things withmanage.py qmemory
andmanage.py qmonitor
.
Thanks, helpful.

roughnecks via Mailman-users writes:
if you look at this graph you'll notice how my VPS "reacted" after mailman3 install.
Looks right to me. Mailman 3 was designed at a time when it was reasonable to expect that a serious list host would be actual hardware, configured pretty much exclusively for the use of the MTA, an httpd, an RDBMS, and the Mailman suite on at least consumer-grade hardware (at that time one-core CPU, 4GB RAM, 25+GB HDD). Sooo ...
By default Mailman core will be running about 15 processes, Postorius about 3, HyperKitty up to about 8 I think, the webserver will be running a bunch, if your MTA is Postfix that will be running a bunch of daemons and firing up worker processes whenever there's something to dov, you'll probably be running a bunch of ancillary daemons for mail (opendkim, Spamassassin, clamav, Amavis) and your RDBMS will be running a couple. Mailman's queue runners (about half of its processes) all wake up frequently and check for work. Janitorial processes like crond and logrotate will have more to do (even if all they do is check that there's nothing to do). You'll be responding to the Internet on multiple ports (most likely all of 22, 25, 80, 443, and 587, maybe a couple more) which will attract the usual conglomeration of Internet busybodies and felons unless you're more diligent than I am about "iptables -A -s x.y.z.w -j DROP" for miscreant x.y.z.w :-).
Somebody posted a couple months back that they were working on a fork of Mailman 3 that would use threads rather than processes as much as possible, but that sounds quite hard (so we're unlikely to do it ourselves) and I haven't heard anything since. And although that would definitely help get LOADAVG down, it wouldn't help that much with %CPU since it doesn't do anything to deal with the busybodies and miscreants pounding on your open ports.
-- GNU Mailman consultant (installation, migration, customization) Sirius Open Source https://www.siriusopensource.com/ Software systems consulting in Europe, North America, and Japan
participants (4)
-
Gerald Vogt
-
Mark Sapiro
-
roughnecks
-
Stephen J. Turnbull