Repository: kafka Updated Branches: refs/heads/0.9.0 ca24692d1 -> a412d8a23
MINOR: Update "Java Version" section Use the latest information from LinkedIn and mention that the latest released version is recommended from a security perspective. Author: Ismael Juma <[email protected]> Reviewers: Grant Henke Closes #603 from ijuma/java-version-recommendation Project: http://git-wip-us.apache.org/repos/asf/kafka/repo Commit: http://git-wip-us.apache.org/repos/asf/kafka/commit/a412d8a2 Tree: http://git-wip-us.apache.org/repos/asf/kafka/tree/a412d8a2 Diff: http://git-wip-us.apache.org/repos/asf/kafka/diff/a412d8a2 Branch: refs/heads/0.9.0 Commit: a412d8a23859928a35a371d21cb1cb585672796f Parents: ca24692 Author: Ismael Juma <[email protected]> Authored: Mon Nov 30 13:57:54 2015 -0800 Committer: Guozhang Wang <[email protected]> Committed: Tue Dec 1 11:14:51 2015 -0800 ---------------------------------------------------------------------- docs/ops.html | 17 ++++++++++------- 1 file changed, 10 insertions(+), 7 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/kafka/blob/a412d8a2/docs/ops.html ---------------------------------------------------------------------- diff --git a/docs/ops.html b/docs/ops.html index 2624527..623ab0f 100644 --- a/docs/ops.html +++ b/docs/ops.html @@ -390,19 +390,22 @@ producer.purgatory.purge.interval.requests=100 Our client configuration varies a fair amount between different use cases. <h3><a id="java" href="#java">Java Version</a></h3> -We're currently running JDK 1.7 u51, and we've switched over to the G1 collector. If you do this (and we highly recommend it), make sure you're 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: +From a security perspective, we recommend you use the latest released version of JDK 1.8 as older freely available versions have disclosed security vulnerabilities. + +LinkedIn is currently running JDK 1.8 u5 (looking to upgrade to a newer version) with the G1 collector. If you decide to use the G1 collector (the current default) and you are still on JDK 1.7, make sure you are on u51 or newer. LinkedIn tried out u21 in testing, but they had a number of problems with the GC implementation in that version. + +LinkedIn's tuning looks like this: <pre> -Xms4g -Xmx4g -XX:PermSize=48m -XX:MaxPermSize=48m -XX:+UseG1GC --XX:MaxGCPauseMillis=20 -XX:InitiatingHeapOccupancyPercent=35 +-XX:MaxGCPauseMillis=20 -XX:InitiatingHeapOccupancyPercent=35 -XX:G1HeapRegionSize=16M </pre> For reference, here are the stats on one of LinkedIn's busiest clusters (at peak): - - 15 brokers - - 15.5k partitions (replication factor 2) - - 400k messages/sec in - - 70 MB/sec inbound, 400 MB/sec+ outbound + - 60 brokers + - 50k partitions (replication factor 2) + - 800k messages/sec in + - 300 MB/sec inbound, 1 GB/sec+ outbound The tuning looks fairly aggressive, but all of the brokers in that cluster have a 90% GC pause time of about 21ms, and they're doing less than 1 young GC per second.
