Red Hat Bugzilla – Bug 1322001
Increase inotify_user_watches to allow for more pods on an atomic-host
Last modified: 2016-11-04 03:27:15 EDT
Description of problem: This bug is in reference to https://github.com/openshift/origin/issues/8209 In a scale-up test of openshift, I noted running out of inotify watches at around 800 pods per node. While this is not currently supported, we are targeting this level soon, and don't want to lose track. There's no danger to increasing this by a factor of 8x.
diff --git a/profiles/atomic-guest/tuned.conf b/profiles/atomic-guest/tuned.conf index 6691f1d..c5a9417 100644 --- a/profiles/atomic-guest/tuned.conf +++ b/profiles/atomic-guest/tuned.conf @@ -15,3 +15,4 @@ nf_conntrack_hashsize=131072 [sysctl] kernel.pid_max=131072 net.netfilter.nf_conntrack_max=1048576 +fs.inotify.max_user_watches=65536 diff --git a/profiles/atomic-host/tuned.conf b/profiles/atomic-host/tuned.conf index 0010077..968366f 100644 --- a/profiles/atomic-host/tuned.conf +++ b/profiles/atomic-host/tuned.conf @@ -15,3 +15,4 @@ nf_conntrack_hashsize=131072 [sysctl] kernel.pid_max=131072 net.netfilter.nf_conntrack_max=1048576 +fs.inotify.max_user_watches=65536
Thanks for the patch, upstream commit: https://git.fedorahosted.org/cgit/tuned.git/commit/?id=52f2cc4ad7ad319f0a747136247e976dd00a7669
============================================ Verified in: tuned-2.7.1-2.el7.noarch tuned-profiles-atomic-2.7.1-2.el7.noarch PASS ============================================ # tuned-adm profile atomic-guest # sysctl -a | grep inotify fs.inotify.max_queued_events = 16384 fs.inotify.max_user_instances = 128 fs.inotify.max_user_watches = 65536 -> OK # tuned-adm profile balanced # sysctl -a | grep inotify fs.inotify.max_queued_events = 16384 fs.inotify.max_user_instances = 128 fs.inotify.max_user_watches = 8192 # tuned-adm profile atomic-host # sysctl -a | grep inotify fs.inotify.max_queued_events = 16384 fs.inotify.max_user_instances = 128 fs.inotify.max_user_watches = 65536 -> OK # tuned-adm profile realtime # sysctl -a | grep inotify fs.inotify.max_queued_events = 16384 fs.inotify.max_user_instances = 128 fs.inotify.max_user_watches = 8192 # cat /usr/lib/tuned/atomic-guest/tuned.conf | grep inotify fs.inotify.max_user_watches=65536 # cat /usr/lib/tuned/atomic-host/tuned.conf | grep inotify fs.inotify.max_user_watches=65536 ============================================ Reproduced in: tuned-2.5.1-4.el7.noarch tuned-profiles-atomic-2.5.1-4.el7.noarch FAIL ============================================ # tuned-adm profile atomic-guest # sysctl -a | grep inotify fs.inotify.max_queued_events = 16384 fs.inotify.max_user_instances = 128 fs.inotify.max_user_watches = 8192 -> FAIL # tuned-adm profile balanced # sysctl -a | grep inotify fs.inotify.max_queued_events = 16384 fs.inotify.max_user_instances = 128 fs.inotify.max_user_watches = 8192 # tuned-adm profile atomic-host # sysctl -a | grep inotify fs.inotify.max_queued_events = 16384 fs.inotify.max_user_instances = 128 fs.inotify.max_user_watches = 8192 -> FAIL
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://rhn.redhat.com/errata/RHBA-2016-2479.html