We recently upgraded to Jenkins 1.490, where we encountered severe page load
time issues after running stable for months.  We traced these issues to the
lazy page load features, and chose to downgrade until performance of those
features improves.

Unfortuantely, upon downgrading, jenkins performance issues occured in other
ways.  After a fresh restart, Jenkins may remain snappy and performant for
hours, but inevitably at some point, some pages take a long time to load. 
Once that starts, over the next hour, all pages will take a long time to
load, eventually becomming all but unresponsive (up to 2 minute load times).

We are trying to investigate the root cause.  I've used a lot of java
applications and it feels like what happens when there are no free threads
able to process a request.

I honestly don't know where to begin, we've been looking at thread dumps,
heap dumps and statistics from javaMelody (the monitoring plugin) as well as
perfmon on our master server looking for clues as to what could be causing
the issues. 

The machine we're running is a beefy 8 core, 32GB of ram Win 2008R2 server. 
We've seen the same performance issues on versions 1.480, 1.481, 1.484, and
1.485.  Looking at thread dumps I don't perceive any erroneous threads in
wait, I see the executor threads (we run 7 build agents, with 3 executors
each) either working when builds are in process or waiting when not, etc. 
I'm not exactly sure which threads are responsible for the website.

We've never encountered these kinds of issues with Jenkins before after a
year of successful use, but we're at the end of our rope, having to restart
Jenkins at least twice a day causing massive disruption in our CI routines. 

I've tried finding "best practice" documentation for troubleshooting
performance issues, ways in which to identify if a plugin is the issue or
what might be causing problems but I'm coming up dry.  Any help that can be
provided would be great.  It seems like this should be an easy problem to
find root cause of, there has to be some kind of lock contention on some
resource going on.



--
View this message in context: 
http://jenkins.361315.n4.nabble.com/Jenkins-1-480-1-485-performance-Page-loads-progressively-get-slower-until-unuseable-2-minute-load-ti-tp4649103.html
Sent from the Jenkins issues mailing list archive at Nabble.com.

Reply via email to