fluent / fluentd-kubernetes-daemonset

Fluentd daemonset for Kubernetes and it Docker image
Apache License 2.0
1.27k stars 980 forks source link

Permission problem access /var/log/containers/* #61

Closed jfillman closed 6 years ago

jfillman commented 7 years ago

This issue must have come up before so apologies for not finding it.

My fluentd containers are now run as user '1000' and so don't have read access to /var/lib/docker/containers which is owned by root and has 0700 perms.

I get the log error mesg:

unreadable. It is excluded and would be examined next time. I can't seem to figure out how to run my container as root or grant the necessary perms on the volume. Adding a securitycontext didn't work I'm a kubernetes beginner so hand holding is much appreciated. Here's my daemonset: ``` apiVersion: extensions/v1beta1 kind: DaemonSet metadata: name: fluentd namespace: kube-system labels: k8s-app: fluentd-logging version: v1 kubernetes.io/cluster-service: "true" spec: template: metadata: labels: k8s-app: fluentd-logging version: v1 kubernetes.io/cluster-service: "true" spec: tolerations: - key: node-role.kubernetes.io/master effect: NoSchedule securityContext: runAsNonRoot: false runAsUser: 0 containers: - name: fluentd image: /fluentd-syslog-ng securityContext: runAsNonRoot: false runAsUser: 0 imagePullPolicy: Always env: - name: SYSLOG_SERVER value: "SYSLOGSERVER" - name: SYSLOG_PORT value: "516" - name: API_SERVER value: "https://FQDN" resources: limits: memory: 200Mi requests: cpu: 100m memory: 200Mi volumeMounts: - name: vartmp mountPath: /var/tmp - name: varlog mountPath: /var/log - name: varlibdockercontainers mountPath: /var/lib/docker/containers readOnly: true terminationGracePeriodSeconds: 30 volumes: - name: vartmp hostPath: path: /var/tmp - name: varlog hostPath: path: /var/log - name: varlibdockercontainers hostPath: path: /var/lib/docker/containers ```
flowstate commented 6 years ago

This is also happening to me. I can run bash on the box as the same user (root) and cat all the logs. But I am also getting the 'log unreadable' error. Any chance we can get some help? It's really disappointing that there's been no feedback on this issue for 28 days.

pirankashani commented 6 years ago

Hi, I had same problem and solved when I delete --lod-driver=journald in docker.conf file. If you check the folder /var/log/containers it would be empty and that's why you receive this error message.

mvernimmen commented 6 years ago

Could you try this?

securityContext:
   readOnlyRootFilesystem: true
edsiper commented 6 years ago

Considering this ticket as solved due to previous answers.

jiuchongxiao commented 6 years ago

@jfillman did you solved it? and how solved? i have the same issue.

xinghui-hou commented 5 years ago

I have same problem . flunt-bit output log

[2019/10/11 10:43:59] [error] [in_tail] read error, check permissions: /home/docker/containers/*/*.log check the directory permissions 0755 is ok

gotoworld commented 4 years ago

same issue

[2020/02/26 03:53:13] [error] [in_tail] read error, check permissions: /data/ratelimit-logs/gateway_2.log [2020/02/26 03:53:13] [error] [in_tail] read error, check permissions: /data/ratelimit-logs/gateway_ratelimit.log

repeatedly commented 4 years ago

This is not fluent-bit daemonset repository. If you have a problem with fluent-bit, post it to correct place.

jpiper commented 3 years ago

I had this problem and it was due to SELinux profiles

perezmark commented 1 year ago

@jpiper how did you solve it?

jpiper commented 3 months ago

I just set SELinux to permissive. Not really a solution, just a workaround

nhenderson commented 2 months ago

@jpiper and @perezmark

Updating the containers securityContext and adding the spc_t SELinux policy to the fluentd DaemonSet worked for me:

Like so:

containers:
- name: fluentd
  securityContext:
    runAsNonRoot: false
    runAsUser: 0
    seLinuxOptions:
      type: spc_t

This may be a better solution than just turning off SELinux on the all the nodes.