From f0de8e755b0739acc2804e796f90591cff9bfebe Mon Sep 17 00:00:00 2001 From: Chris Seto Date: Thu, 4 May 2017 13:18:23 -0400 Subject: [PATCH 1/2] Add note about container based environments --- 510_Deployment/20_hardware.asciidoc | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/510_Deployment/20_hardware.asciidoc b/510_Deployment/20_hardware.asciidoc index ec3954462..24ad13f61 100644 --- a/510_Deployment/20_hardware.asciidoc +++ b/510_Deployment/20_hardware.asciidoc @@ -33,6 +33,13 @@ If you need to choose between faster CPUs or more cores, choose more cores. The extra concurrency that multiple cores offers will far outweigh a slightly faster clock speed. +If you are running in a container based environment, such as kubernetes, that would +result in multiple elasticsearch instances running on a single host, make sure you set +the `processors` settings appropriately. Each instance will report the full number of cores avaiable. +A cluster with two instances running on a 32-core machine will think it has 64-cores total. This will +result in over allocation of threadpools, causing the nodes to use more memory than the have +the resources to garbage collect efficiently. + ==== Disks Disks are important for all clusters,((("disks")))((("hardware", "disks"))) and doubly so for indexing-heavy clusters From 18161d13044405652106190b9f650337128d253a Mon Sep 17 00:00:00 2001 From: Chris Seto Date: Thu, 4 May 2017 16:44:10 -0400 Subject: [PATCH 2/2] Update 20_hardware.asciidoc --- 510_Deployment/20_hardware.asciidoc | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/510_Deployment/20_hardware.asciidoc b/510_Deployment/20_hardware.asciidoc index 24ad13f61..7b3ff1c7c 100644 --- a/510_Deployment/20_hardware.asciidoc +++ b/510_Deployment/20_hardware.asciidoc @@ -37,8 +37,8 @@ If you are running in a container based environment, such as kubernetes, that wo result in multiple elasticsearch instances running on a single host, make sure you set the `processors` settings appropriately. Each instance will report the full number of cores avaiable. A cluster with two instances running on a 32-core machine will think it has 64-cores total. This will -result in over allocation of threadpools, causing the nodes to use more memory than the have -the resources to garbage collect efficiently. +result in over allocation of threadpools, causing the instance to use more memory than the have +the resources to garbage collect efficiently. Eventually leading to instances getting evicted from the cluster. ==== Disks