Created attachment 1565354 [details] Tuned default Description of problem: Values from tuned default doesn't apply. Version-Release number of selected component (if applicable): $ oc get clusterversion NAME VERSION AVAILABLE PROGRESSING SINCE STATUS version 4.1.0-0.nightly-2019-05-07-155635 True False 30m Cluster version is 4.1.0-0.nightly-2019-05-07-155635 $ oc get co/node-tuning NAME VERSION AVAILABLE PROGRESSING DEGRADED SINCE node-tuning 4.1.0-0.nightly-2019-05-07-155635 True False False 40m How reproducible: 100% Steps to Reproduce: 1. Default next-gen installation. 2. Edit default values: $ oc edit tuned default -n openshift-cluster-node-tuning-operator ex. ... net.netfilter.nf_conntrack_max=1048578 ... 3. Save changes, check value on every node: $ for node in $(oc get nodes --no-headers | cut -d ' ' -f 1); do oc debug node/$node -- chroot /host && sysctl net.netfilter.nf_conntrack_max; done Actual results: net.netfilter.nf_conntrack_max = 262144 Expected results: the same like in tuned default: net.netfilter.nf_conntrack_max=1048578
Created attachment 1565355 [details] Pod log
You're checking the sysctl on your desktop/wherever you're running your "oc debug node" command from. You need to drop the "&&" in your "oc debug node" one-liner. I cannot reproduce the issue on exactly the same nightly: $ oc get clusterversion NAME VERSION AVAILABLE PROGRESSING SINCE STATUS version 4.1.0-0.nightly-2019-05-07-155635 True False 9m27s Cluster version is 4.1.0-0.nightly-2019-05-07-155635 $ oc get nodes NAME STATUS ROLES AGE VERSION ip-10-0-130-239.us-west-2.compute.internal Ready master 20m v1.13.4+0af5e0f8b ip-10-0-143-89.us-west-2.compute.internal Ready worker 13m v1.13.4+0af5e0f8b ip-10-0-158-40.us-west-2.compute.internal Ready worker 12m v1.13.4+0af5e0f8b ip-10-0-181-151.us-west-2.compute.internal Ready worker 13m v1.13.4+0af5e0f8b $ oc project openshift-cluster-node-tuning-operator Now using project "openshift-cluster-node-tuning-operator" on server "https://api.jmencak.perf-testing.devcluster.openshift.com:6443". $ oc edit tuned/default tuned.tuned.openshift.io/default edited $ date Wed 08 May 2019 09:14:38 AM CEST $ oc get pods -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES cluster-node-tuning-operator-54f64947f5-g7fg9 1/1 Running 0 17m 10.128.0.44 ip-10-0-130-239.us-west-2.compute.internal <none> <none> tuned-c4vxk 1/1 Running 0 14m 10.0.158.40 ip-10-0-158-40.us-west-2.compute.internal <none> <none> tuned-ksv2v 1/1 Running 0 14m 10.0.181.151 ip-10-0-181-151.us-west-2.compute.internal <none> <none> tuned-n7wrx 1/1 Running 0 16m 10.0.130.239 ip-10-0-130-239.us-west-2.compute.internal <none> <none> tuned-s48z5 1/1 Running 0 14m 10.0.143.89 ip-10-0-143-89.us-west-2.compute.internal <none> <none> $ oc logs tuned-c4vxk | tail -n20 I0508 07:14:47.069790 1667 openshift-tuned.go:638] Remove event on: /var/lib/tuned/profiles-data/..2019_05_08_07_00_18.128954904 I0508 07:14:48.589697 1667 openshift-tuned.go:176] Extracting tuned profiles I0508 07:14:48.590533 1667 openshift-tuned.go:215] Reloading tuned... 2019-05-08 07:14:48,681 INFO tuned.daemon.application: dynamic tuning is globally disabled 2019-05-08 07:14:48,686 INFO tuned.daemon.daemon: using sleep interval of 1 second(s) 2019-05-08 07:14:48,686 INFO tuned.daemon.daemon: Running in automatic mode, checking what profile is recommended for your configuration. 2019-05-08 07:14:48,688 INFO tuned.daemon.daemon: Using 'openshift-node' profile 2019-05-08 07:14:48,688 INFO tuned.profiles.loader: loading profile: openshift-node 2019-05-08 07:14:48,723 WARNING tuned.daemon.application: Using one shot no deamon mode, most of the functionality will be not available, it can be changed in global config 2019-05-08 07:14:48,723 INFO tuned.daemon.controller: starting controller 2019-05-08 07:14:48,723 INFO tuned.daemon.daemon: starting tuning 2019-05-08 07:14:48,725 INFO tuned.daemon.controller: terminating controller 2019-05-08 07:14:48,725 INFO tuned.daemon.daemon: stopping tuning 2019-05-08 07:14:48,732 INFO tuned.plugins.base: instance cpu: assigning devices cpu0, cpu1 2019-05-08 07:14:48,735 WARNING tuned.plugins.plugin_cpu: your CPU doesn't support MSR_IA32_ENERGY_PERF_BIAS, ignoring CPU energy performance bias 2019-05-08 07:14:48,736 INFO tuned.plugins.base: instance disk: assigning devices xvda 2019-05-08 07:14:48,736 INFO tuned.plugins.base: instance net: assigning devices ens3 2019-05-08 07:14:48,999 INFO tuned.plugins.plugin_sysctl: reapplying system sysctl 2019-05-08 07:14:49,004 INFO tuned.daemon.daemon: static tuning from profile 'openshift-node' applied 2019-05-08 07:14:49,012 INFO tuned.daemon.daemon: terminating Tuned $ for node in $(oc get nodes --no-headers | cut -d ' ' -f 1); do oc debug node/$node -- chroot /host sysctl net.netfilter.nf_conntrack_max; done Starting pod/ip-10-0-130-239us-west-2computeinternal-debug ... To use host binaries, run `chroot /host` net.netfilter.nf_conntrack_max = 1048578 Removing debug pod ... Starting pod/ip-10-0-143-89us-west-2computeinternal-debug ... To use host binaries, run `chroot /host` net.netfilter.nf_conntrack_max = 1048578 Removing debug pod ... Starting pod/ip-10-0-158-40us-west-2computeinternal-debug ... To use host binaries, run `chroot /host` net.netfilter.nf_conntrack_max = 1048578 Removing debug pod ... Starting pod/ip-10-0-181-151us-west-2computeinternal-debug ... To use host binaries, run `chroot /host` net.netfilter.nf_conntrack_max = 1048578 Removing debug pod ... Also, please note that: 1) tuned-profiles/tuned-recommend ConfigMap change propagation to tuned pods might take a while 2) the default tuned CR is not really meant for user changes and you should use custom tuned CRs instead Closing the bug, please re-open if the solution doesn't work for you.
http://www.reactiongifs.com/r/whid.gif you have absolutely right! My bad. Sorry for taking your time.