Just about all of our systems are actually running 64GB of memory now, and yes, we want plenty of space for cache with all the disk and network IO going on.
-Todd On 5/7/14, 10:31 AM, "John Penner" <j...@longtailvideo.com> wrote: >Just curious if these JAVA_HEAP tunings are still for machines listed in >6.4 >Hardware and OS <http://kafka.apache.org/documentation.html#hwandos>: >"We are using dual quad-core Intel Xeon machines with 24GB of memory." > >Seems like a small heap size for 24GB of memory, or is that done to >maximize as much memory as possible for pagecache? > >-penner > > >On Tue, May 6, 2014 at 1:50 PM, Neha Narkhede ><neha.narkh...@gmail.com>wrote: > >> Updated the docs with Todd's recommendation - >> http://kafka.apache.org/081/documentation.html#java >> >> Thanks, >> Neha >> >> >> On Tue, May 6, 2014 at 9:40 AM, Jay Kreps <jay.kr...@gmail.com> wrote: >> >> > Hey Todd, >> > >> > Doc patch? :-) >> > >> > svn co http://svn.apache.org/repos/asf/kafka/site/081/ >> > >> > Don't stress about html or formatting, I'm happy to do that part. I >>would >> > love to give people more authoritative advice. Right now everything >>is a >> > bit obsolete and wrong. >> > >> > -Jay >> > >> > >> > On Mon, May 5, 2014 at 10:36 PM, Todd Palino <tpal...@linkedin.com> >> wrote: >> > >> > > I apologize for taking a couple days to jump in on this. We¹re >> currently >> > > running JDK 1.7 u51, and we¹ve switched over to the G1 collector. If >> you >> > > do this (and I highly recommend it), make sure you¹re on u51. We >>tried >> > out >> > > u21 in testing, but we had a number of problems with the GC >> > implementation >> > > in that version. >> > > >> > > Our tuning looks like this: >> > > -Xms4g -Xmx4g -XX:PermSize=48m -XX:MaxPermSize=48m -XX:+UseG1GC >> > > -XX:MaxGCPauseMillis=20 -XX:InitiatingHeapOccupancyPercent=35 >> > > >> > > For reference, I¹m looking at the stats on one of our busiest >>clusters >> > (at >> > > peak): >> > > - 15 brokers >> > > - 15.5k partitions (replication factor 2) >> > > >> > > - 400k messages/sec in >> > > - 70 MB/sec inbound, 400 MB/sec+ outbound >> > > >> > > The tuning looks fairly aggressive, but all of our brokers in that >> > cluster >> > > have a 90% GC pause time of about 21ms, and they¹re doing less than >>1 >> > > young GC per second. We haven¹t seen a single full GC on those >>brokers >> in >> > > the last month, and previous to that I think we only saw them when I >> was >> > > messing around with the cluster in a very painful way, not under >> anything >> > > approaching normal traffic. >> > > >> > > -Todd >> > > >> > > >> > > On 5/1/14, 9:21 PM, "Neha Narkhede" <neha.narkh...@gmail.com> wrote: >> > > >> > > >The GC settings at >>http://kafka.apache.org/documentation.html#javaare >> > > >old. >> > > >We meant to update the documentation with the new GC settings using >> the >> > G7 >> > > >collector, but we haven't gotten around to doing that. Let me reach >> out >> > to >> > > >our engineer, Todd Palino, who worked on tuning GC for Kafka at >> LinkedIn >> > > >to >> > > >see if we can update our docs. >> > > > >> > > >Thanks, >> > > >Neha >> > > > >> > > > >> > > >On Thu, May 1, 2014 at 9:02 PM, Jun Rao <jun...@gmail.com> wrote: >> > > > >> > > >> http://kafka.apache.org/documentation.html#java >> > > >> >> > > >> Thanks, >> > > >> >> > > >> Jun >> > > >> >> > > >> >> > > >> On Thu, May 1, 2014 at 12:19 PM, Cassa L <lcas...@gmail.com> >>wrote: >> > > >> >> > > >> > Hi, >> > > >> > I want to know what usually, are the JAVA_HEAP settings >> > recommended >> > > >>for >> > > >> > kafka servers in production. >> > > >> > >> > > >> > Thanks >> > > >> > LCassa >> > > >> > >> > > >> >> > > >> > > >> > >>
