From c0c775b4e2e5debcf86c4eb4834030596f673c41 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Marcin=20=27Perk=27=20Sto=C5=BCek?= Date: Fri, 2 Apr 2021 15:45:49 +0200 Subject: [PATCH] Add troubleshooting section for the stuck PVCs --- deploy/docs/Troubleshoot_Collection.md | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) diff --git a/deploy/docs/Troubleshoot_Collection.md b/deploy/docs/Troubleshoot_Collection.md index 8c2ef75283..654fed620b 100644 --- a/deploy/docs/Troubleshoot_Collection.md +++ b/deploy/docs/Troubleshoot_Collection.md @@ -16,6 +16,7 @@ - [Common Issues](#common-issues) - [Missing metrics - cannot see cluster in Explore](#missing-metrics---cannot-see-cluster-in-explore) - [Pod stuck in `ContainerCreating` state](#pod-stuck-in-containercreating-state) + - [Fluentd Pod stuck in `Pending` state after recreation](#fluentd-pod-stuck-in-pending-state-after-recreation) - [Missing `kubelet` metrics](#missing-kubelet-metrics) - [1. Enable the `authenticationTokenWebhook` flag in the cluster](#1-enable-the-authenticationtokenwebhook-flag-in-the-cluster) - [2. Disable the `kubelet.serviceMonitor.https` flag in Kube Prometheus Stack](#2-disable-the-kubeletservicemonitorhttps-flag-in-kube-prometheus-stack) @@ -260,6 +261,21 @@ Warning FailedCreatePodSandBox 29s kubelet, ip-172-20-87-45.us-west-1.comput you have an unhealthy node. Killing the node should resolve this issue. +### Fluentd Pod stuck in `Pending` state after recreation + +If you are seeing a Fluentd Pod stuck in the `Pending` state, using the [file based buffering](./Best_Practices.md#fluentd-file-based-buffer) +(default since v2.0) and seeing logs like + +``` +Warning FailedScheduling 16s (x23 over 31m) default-scheduler 0/6 nodes are available: 2 node(s) had volume node affinity conflict, 4 node(s) were unschedulable. +``` + +you have a volume node affinity conflict. It can happen when Fluentd Pod was running in one AZ and has been rescheduled into +another AZ. Deleting the existing PVC and then killing the Pod should resolve this issue. + +The Fluentd StatefulSet Pods and their PVCs are bound by their number: `*-sumologic-fluentd-logs-1` Pod will be using +the `buffer-*-sumologic-fluentd-logs-1` PVC. + ### Missing `kubelet` metrics Navigate to the `kubelet` targets using the steps above. You may see that the targets are down with 401 errors. If so, there are two known workarounds you can try.