> On 31 Aug 2017, at 17:55, Christopher Schultz <ch...@christopherschultz.net> > wrote: > > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA256 > > Andrea, > > On 8/30/17 4:13 PM, Ing. Andrea Vettori wrote: >>> On 30 Aug 2017, at 00:16, Christopher Schultz >>> <ch...@christopherschultz.net> wrote: RMI is known for flagrantly >>> wasting permgen/metaspace because of all the Proxy objects and >>> special throw-away Class objects it uses, etc. But I'm not sure >>> why the metaspace filling up would have such a dramatic effect on >>> performance. If you were stalling during a metaspace collection, >>> it might make sense, but that's not what you are reporting. >> >> That’s true in fact what I’d like to understand is if a big >> metaspace can affect performance… one thing I want to try is to run >> the app with metaspace capped at 400/500MB. On java 7 I had PermGen >> at 512 and as far as I remember, it was working fine. > > Yes, I think limiting the metaspace might help. It may result in more > frequent, smaller GC cycles. > > I actually don't know much about the (relatively) new metaspace > region, but it's possible that GC in the metaspace is a "stop the > world" operation where the JVM can't do anything else during that time.
I tried that this morning. I set one tomcat with max metaspace set to 512mb, left the other one with 2gb. The one with 512mb was faster to call jboss on average but triggered a lot more the gc so some calls where very slow (the ones when the gc triggered). If you’ve seen my other reply the performance problem has been fixed but the curiosity to understand why big metaspace is slow remains… Thanks --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
