05:25 <GitHub69> [jruby] bjfish pushed 1 new commit to truffle-head: https://git.io/viP8z

05:25 <GitHub69> jruby/truffle-head 41e28fd Brandon Fish: [Truffle] String#slice use :to_int for conversion

16:03 <sam2000_> Hi Guys, I am trying to do some Jruby memory profiling for a rails up. I tried using VisualVM and found there is an increasing number of instances of some java collection classes and I am not sure where they are coming from. Java. Util.HashMap nodes and Hashmaps top the list. Any ideas please?

16:09 <sam2000_> To add some more details - I am using Jruby 9.1.2 with rails 4 - I am not using any global variables. This things is really driving me nuts so I would really appreciate any help please

18:44 <chrisseaton> JRuby uses those classes internally for many things

18:44 <chrisseaton> I'm not sure how you'd pin it down further

18:44 <chrisseaton> Are you doing anything like dynamically creating lots of classes?

19:51 <sam2000> Thanks very much for reply @chrisseaton. If it is used by Jruby internally then Maybe I shouldn’t worry about it. When I profiled I noticed that the number of instances seem to increase with requests. At the moment it is around 200,000 and it increases by about 20,000 after a few requests. But it doesn’t seem to take a lot of memory though. The heap size looks pretty much flat. Although you can see a tiny increase if yo

19:51 <sam2000> look at the exact numbers

19:52 <chrisseaton> If you are allocating new hash map nodes while handling requests, after everyone is loaded, that may be indicative that something isn't right

19:52 <chrisseaton> Such as classes being created dynamically

19:52 <chrisseaton> After everything is up and running I can't see why you'd need new hash map nodes

19:54 <sam2000> So you think there must be a problem somewhere then? I am wondering what would be the next step in getting closer to the problem? Could you suggest anything please?

19:55 <chrisseaton> Are you able to say how many hash map nodes are being allocated per second?

19:55 <chrisseaton> Maybe give the app 3 minutes to warm up, put it under heavy load, and then see how many are allocated

19:55 <chrisseaton> Then open an issue if it's more than a few hundred per second or something

19:55 <chrisseaton> Ping me on it

20:03 <sam2000> I have just looked at the 3 heap dumps i have taken recently and it has increased by 35000 in five imnutes and then increased by about 26000 in about 20 minutes. I can do some more testing but do these numbers look normal ?

20:03 <sam2000> The numbers are the number of hasmap node instances

20:07 <sam2000> By the way these heap dumps are taken with very little load, just a few requests I made by hand using a browser, not more than 30 - 40 requests between dumps

21:19 <sam2000> Thanks @chrisseaton I did some test and after the initial rise after start up, it is about 70 per second. From what you were saying I am guessing this is to be expected?