Hmm, maybe thats myfaces update. I have an app where starting using jsf (f5) simply make my mem x2.
Can you check downgrading mf please? Le 2 mai 2013 07:03, "Howard W. Smith, Jr." <[email protected]> a écrit : > Okay, I will have to try that, ASAP, thanks...and you're welcome. :) > > On a tangent, I have had the following java memory/GC settings (at least 1 > or 2 months now), > > -Xms1250m > -Xmx1250m > -XX:MaxPermSize=384m > -XX:+UseTLAB > -XX:+UseConcMarkSweepGC > -XX:+CMSClassUnloadingEnabled > > Before using tomee 1.6.0 snapshot (that has owb 1.2), I would open > jvisualvm on production server, monitor GC and memory used, and I would see > memory used between 120m and 400m. > > Now, since I'm using tomee 1.6.0 snapshot (with owb 1.2), memory levels > seem to be higher; on average, I think I'm seeing between 400m and 600m. > > I have been using tomee 1.6.0 snapshot (with owb 1.2) in production for 1 > or 2 days now. About 1 hour ago, I deployed a new WAR (with some code > changes of my own), and I did not do much in the app at all, just logged > in, and logged out. so, an hour later and even now, I'm seeing between 500m > and 600m. > > what is owb 1.2 doing, now, that could be causing higher memory levels. I > will tell you that even when no users are connected to the server, I do > have @Schedule methods that are triggered every 30 seconds (checking the > value of some CDI @ApplicationScoped bean attributes/members) and every 2 > minutes, checking an email server. > > Earlier (before owb 1.2), when no users connected to the app, TomEE would > have memory used right around 200m (on average). that is why I set > Xms/Xmx1250m (per tomcat's Mark Thomas; 5 x 250m, since 250m is least > amount of memory used..on average). > > bit of a background... I used tomee 1.6.0 snapshot (2013-03-27) all > throughout April 2013 until I recently started using tomee 1.6.0 snapshot > (with owb 1.2). > > > > On Thu, May 2, 2013 at 12:38 AM, Romain Manni-Bucau > <[email protected]>wrote: > > > Thanks ;) > > > > About exceptions: start without breakpoints then put it in > > java.lang.Exception constructors then hit f5. It will be easier IMO. > > Le 2 mai 2013 01:35, "Howard W. Smith, Jr." <[email protected]> a > > écrit : > > > > > Mark, > > > > > > Per Romain's request, I thought I might change the subject. see my > > > responses below. > > > > > > ---------- Forwarded message ---------- > > > > > > > > > > > >>a.) There should be no need to change anything for you in your own > > code. > > > All the refactoring is just internally. > > > >> > > > >> > > > > > > > > > > > >sorry, don't understand. should be no need to change anything in my > > code, > > > and then you say, 'refactoring is just internally'? > > > > > > > > > > The OWB-1.2 changes are purely owb internally. You don't need to > change > > > anything in YOUR code ;) > > > > > > understood, thanks. > > > > > > > > > >b.) If you do performance testing, then please always switch MyFaces > to > > > ProjectStage 'Production'. > > > >> > > > > > > > > > > > >done... i keep project to 'production'...always. I learned that by the > > big > > > 'comment' outputted in server log when I started using MyFaces (many > > months > > > ago). :) > > > > > > > > > > Nah, usually 'Development' is perfectly fine. We do a lot of caching > > only > > > in Production, but it's disabled for 'Development'. In Dev mode we e.g. > > > always read the xhtml pages from the files for each request and build > up > > > the facelets tree info. Whereas in Production we only read the xhtml > > once. > > > But then you would not be able to change your xhtml files on the fly in > > dev > > > mode... > > > > > > Interesting. I will have to keep this in mind, and this motivates me to > > > change project state to development and do some testing. I'm even > > wondering > > > what will happen if I revert to MyFaces 2.1.10 with latest tomee 1.6.0 > > > snapshot (without touching/removing owb 1.2 JARs), and a few other > > > tomee/myfaces/atmosphere configurations. > > > > > > I will try to do this, ASAP, and report test results, here as well. > > > > > > >c.) regarding atmosphere: you could use the tomcate WebSocket stuff > > which > > > is more EE7 like already. > > > >> > > > > > > > > > > > >interesting... i have seen some on tomcat list ask about tomcat > > websocket > > > stuff, and I have recognized that tomcat has did some websocket changes > > in > > > tomcat 7.0.35 or 7.0.37 release (forgot which one, exactly). > > > > > > > > > > > >are you saying that I could replace atmosphere with tomcat websocket > > > stuff? if so, wow... free myself of atmosphere 'dependency'?!?! :) > > > > > > > > > > That was before I did read you actually use primefaces push which > > really > > > needs atmosphere. > > > > > > okay. > > > > > > > > > >when I was using glassfish, I saw websocket/comet samples in the past, > > but > > > thought it was 'easier' to use PrimeFaces Push (and that was the first > > > reason why I started using tomee/tomcat7, since primefaces push or > > > atmosphere needs a lot of work to work with glassfish). > > > > > > > > > > > > > > > >>d.) OWB does not yet support async-supported. But having 1 element in > > the > > > servlet chain which doesnt set this flag makes the servlet container > > switch > > > off that feature anyway. > > > >> > > > > > > > > > > > >hmm interesting... i am aware that servlet 3 has async support; seems > as > > > though you are telling me that OWB is not-yet-fully > servlet-3-compliant, > > > but I'm not trying to say that too loudly, so please feel free to > correct > > > me. :) > > > > > > > > > > OpenWebBeans IS fully servlet-3 compliant. But it does not use the > > > async-spported feature yet. > > > > > > okay. > > > > > > > > > > > > > > > > >also, in response to you, I am not implicitly setting async-supported > > > anywhere in my web.xml. > > > > > > > > > > > > > > > > > > > >There are 2 variables in the game > > > >>1.) tomee > > > >>2.) atmosphere > > > >> > > > >>if you switch back to an older tomee version: works > > > >> > > > > > > > > > > > >correct; from what I remember, an older version of tomee (owb 1.1.8) > > works > > > with latest version of atmosphere > > > > > > > > > > > >if you keep newest tomee and only switch back atmosphere to an older > > > version: works > > > >> > > > > > > > > > > > >correct; newest tomee works with older atmosphere > > > > > > > >only if you have both newest tomee and newest atmosphere: broken > > > >> > > > > > > > > > > > >almost-correct. :) > > > > > > > > > > > >both atmosphere and tomee are telling me that this is neither an issue > > > with atmosphere or tomee, so i don't want to say 'broken'; the only > > > unwanted behavior is the 'new' atmosphere log line that shows up in my > > > server/tomcat7-stderr/catalina log on every full page refresh :) > > > > > > > > > > > > > > > > > > > >>Oki, so now what does all the logs tell you? > > > > > > > > > > > >Per the test case at the following: > > > > > > > > > > > >https://github.com/smithh032772/PrimePushCounter.git > > > > > > > > > > > > > > > >the log shows the following whenever I press F5 key to refresh page in > > > browser: > > > > > > > > > > > >Apr 30, 2013 4:59:15 PM org.atmosphere.cpr.DefaultBroadcaster <init> > > > >INFO: /counter support Out Of Order Broadcast: false > > > >Apr 30, 2013 5:02:40 PM org.atmosphere.cpr.DefaultBroadcaster <init> > > > >INFO: /counter support Out Of Order Broadcast: false > > > > > > > > > > > > > > > >Any Exceptions which got swallowed, etc? > > > >> > > > > > > > > > > > >very good question. to this day, i have not learned how to recover > > > tomcat-swallowed exceptions. :) > > > > > > Start up your debugger and let it break on any exception. Then exclude > > > stuff like java.io.* etc until you find something you which looks > > curious. > > > > > > hmmm... i understand 'let it break on any exception', but exclude stuff > > > like java.io.* ? > > > > > > > > > >i hope this response helps. > > > > > > > > > > > > > > > > > > > > > >
