I'm running a Jenkins instance with ~7000 jobs (that is otherwise
unremarkable). The limiting factor with many jobs is mostly plugins or specific
core features that have scalability problems. Examples:
* Cloudbees RBAC had really slow group membership resolutions a few versions
ago, leading to extremely long page load times. It's much improved now.
* Subversion plugin often sends an unnecessary HTTPS request per module
location per project, making post-commit hooks take forever with a lot of
projects and using up all connections to our SVN server.
* JENKINS-20052 doubles or triples the list view render times, but it doesn't
seem to be a big issue on my instance.
* JENKINS-18660 The 'label cloud' made the slave index page unusable with a lot
of jobs. Resolved in 1.509.3.
In general, you'll notice lazy-loading build loading, e.g. when determining the
combined weather of a folder, or in views with View Job Filters plugin's
filters examining past builds. Make sure you have enough RAM so it only happens
once after a restart (or use a different retention strategy as of 1.548).
---
There's also the anonymous usage statistics that can provide some insight into
what size Jenkins installations exist.
For example, as of December, these are the largest installs by number of nodes
(200+ nodes):
version|jobs|executors|nodes|plugins
1.532.1|38|328|249|31
1.491|3691|699|673|40
1.534|871|263|205|57
1.480.3|72|1084|409|124
1.501|174|704|240|85
1.480.2|1680|671|364|67
1.492|359|520|255|42
1.480.3|99|851|230|113
1.538|20|282|282|15
1.509.2|5642|433|263|123
1.509.2|111|337|258|14
1.480.1|2157|780|382|68
1.429|843|4866|4617|21
1.517|115|231|228|56
1.498|136|316|219|72
1.466.2|1182|470|339|40
1.480.3|206|724|246|120
1.538|4|846|836|40
The stats could also tell you the OS and which plugins in which versions are
used, but it doesn't tell you anything about the hardware used. Still, it's
useful to see that there are some really large installations. Note that the
apparently low number of jobs in some of these large installations (249 slaves
for 38 jobs?) could be due to a bug in how Jenkins reports job numbers
(JENKINS-21448).
On 19.01.2014, at 00:55, Kirill Evstigneev wrote:
> It would be nice to have a list of large Jenkins installations. With numbers
> - how many jobs and builds they run, number of slaves, hardware specs etc.
> There are (sometimes reasonable) concerns regarding Jenkins performance. So
> such real world evidence will help to address it and to assess Jenkins
> ability to scale.
>
> On last two JUCs some numbers were published about these setups:
> - Yahoo Advertising platform;
> - Sony Moblie
> - Netflix
> - Openstack @ HP
> - Cisco Videoscape (very incomplete)
>
> Does anyone knows about any other examples?
>
> --
> You received this message because you are subscribed to the Google Groups
> "Jenkins Users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to jenkinsci-users+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
--
You received this message because you are subscribed to the Google Groups
"Jenkins Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to jenkinsci-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
