Many of these problem cases are hard to summarize down to a short phrase that adequately describes the problem. Each problem will start with a bulleted list of symptoms. Keep in mind that all symptoms may not apply depending upon the configuration of the Rook. If the majority of the symptoms are seen there is a fair chance you are experiencing that problem.
If after trying the suggestions found on this page and the problem is not resolved, the Rook team is very happy to help you troubleshoot the issues in their Slack channel. Once you have [registered for the Rook Slack](https://rook-slackin.herokuapp.com/), proceed to the General channel to ask for assistance.
One of the first things that should be done is to start the [rook-tools pod](./toolbox.md) as described in the Toolbox section. Once the pod is up and running one can `kubectl exec` into the pod to execute Ceph commands to evaluate that current state of the cluster. Here is a list of commands that can help one get an understanding of the current state.
Of particular note, the first two status commands provide the overall cluster health. The normal state for cluster operations is HEALTH_OK, but will still function when the state is in a HEALTH_WARN state. If you are in a WARN state, then the cluster is in a condition that it may enter the HEALTH_ERROR state at which point *all* disk I/O operations are halted. If a HEALTH_WARN state is observed, then one should take action to prevent the cluster from halting when it enters the HEALTH_ERROR state.
There is literally a ton of Ceph sub-commands to look at and manipulate Ceph objects. Well beyond the scope of a few troubleshooting techniques and there are other sites and documentation sets that deal more with assisting one with troubleshooting a Ceph environment. In addition, there are other helpful hints and some best practices concerning a Ceph environment located in the [Advanced Configuration section](advanced-configuration.md). Of particular note there are scripts for collecting logs and gathering OSD information there.
*`rook-agent` pod is in a `CrashLoopBackOff` status because it cannot deploy its driver on a read-only filesystem: [Flexvolume configuration pre-reqs](./k8s-pre-reqs.md#flexvolume-configuration)
* Persistent Volume and/or Claim are failing to be created and bound: [Volume Creation](#volume-creation)
*`rook-agent` pod is failing to mount and format the volume: [Rook Agent Mounting](#volume-mounting)
* You are using Kubernetes 1.7.x or earlier and the Kubelet has not been restarted after `rook-agent` is in the `Running` status: [Restart Kubelet](#kubelet-restart)
* You are using Kubernetes 1.6.x and the attach-detach controller has not been disabled: [Disable attach-detach controller](./quickstart.md#disable-attacher-detacher-controller)
If you see some of the symptoms above, it's because the requested Rook storage for your pod is not being created and mounted successfully.
In this walkthrough, we will be looking at the wordpress mysql example pod that is failing to start.
To first confirm there is an issue, you can run commands similar to the following and you should see similar output (note that some of it has been omitted for brevity):
1h 1h 3 default-scheduler Warning FailedScheduling PersistentVolumeClaim is not bound: "mysql-pv-claim" (repeated 2 times)
1h 35s 36 kubelet, 172.17.8.101 Warning FailedMount Unable to mount volumes for pod "wordpress-mysql-918363043-50pjr_default(08d14e75-bd99-11e7-bc4c-001c428b9fc8)": timeout expired waiting for volumes to attach/mount for pod "default"/"wordpress-mysql-918363043-50pjr". list of unattached/unmounted volumes=[mysql-persistent-storage]
1h 35s 36 kubelet, 172.17.8.101 Warning FailedSync Error syncing pod
```
To troubleshoot this, let's walk through the volume provisioning steps in order to confirm where the failure is happening.
### Rook Agent Deployment
The `rook-agent` pods are responsible for mapping and mounting the volume from the cluster onto the node that your pod will be running on.
If the `rook-agent` pod is not running then it cannot perform this function.
Below is an example of the `rook-agent` pods failing to get to the `Running` status because they are in a `CrashLoopBackOff` status:
```console
> kubectl -n rook-system get pod
NAME READY STATUS RESTARTS AGE
rook-agent-ct5pj 0/1 CrashLoopBackOff 16 59m
rook-agent-zb6n9 0/1 CrashLoopBackOff 16 59m
rook-operator-2203999069-pmhzn 1/1 Running 0 59m
```
If you see this occurring, you can get more details about why the `rook-agent` pods are continuing to crash with the following command and its sample output:
```console
> kubectl -n rook-system get pod -l app=rook-agent -o jsonpath='{range .items[*]}{.metadata.name}{"\t"}{.status.containerStatuses[0].lastState.terminated.message}{"\n"}{end}'
rook-agent-ct5pj mkdir /usr/libexec/kubernetes: read-only file system
rook-agent-zb6n9 mkdir /usr/libexec/kubernetes: read-only file system
```
From the output above, we can see that the agents were not able to bind mount to `/usr/libexec/kubernetes` on the host they are scheduled to run on.
For some environments, this default path is read-only and therefore a better path must be provided to the agents.
Once the `rook-agent` pods are gone, **follow the instructions in the [Flexvolume configuration pre-reqs](./k8s-pre-reqs.md#flexvolume-configuration)** to ensure a good value for `--volume-plugin-dir` has been provided to the Kubelet.
After that has been configured, and the Kubelet has been restarted, start the agent pods up again by restarting `rook-operator`:
```console
kubectl -n rook-system delete pod -l app=rook-operator
```
### Kubelet Restart
#### **Kubernetes 1.7.x and earlier only**
If the `rook-agent` pods are all in the `Running` state then another thing to confirm is that **if you are running on Kubernetes 1.7.x or earlier**, the Kubelet must be restarted after the `rook-agent` pods are running.
A symptom of this can be found in the Kubelet's log/journal, with the following error saying `no volume plugin matched`:
```console
Oct 30 22:23:03 core-02 kubelet-wrapper[31926]: E1030 22:23:03.524159 31926 desired_state_of_world_populator.go:285] Failed to add volume "mysql-persistent-storage" (specName: "pvc-9f273fbc") for pod "9f2ff89a-bdbf" to desiredStateOfWorld. err=failed to get Plugin from volumeSpec for volume "pvc-9f273fbc" err=no volume plugin matched
If you encounter this, just **restart the Kubelet process**, as described in the **[Restart Kubelet](./quickstart.md#restart-kubelet)** section of the Rook deployment guide.
If the volume is failing to be created, there should be details in the log output, especially those tagged with `op-provisioner`.
### Volume Mounting
The final step in preparing Rook storage for your pod is for the `rook-agent` pod to mount and format it.
If all the preceding sections have been successful or inconclusive, then take a look at the `rook-agent` pod logs for further clues.
You can determine which `rook-agent` is running on the same node that your pod is scheduled on by using the `-o wide` output, then you can get the logs for that `rook-agent` pod similar to the example below:
In the `rook-agent` pod logs, you may see a snippet similar to the following:
```console
2017-11-07 00:04:37.808870 I | rook-flexdriver: WARNING: The node kernel version is 4.4.0-87-generic, which do not support multiple ceph filesystems. The kernel version has to be at least 4.7. If you have multiple ceph filesystems, the result could be inconsistent
```
This will happen in kernels with versions older than 4.7, where the option `mds_namespace` is not supported. This option is used to specify a filesystem namespace.
In this case, if there is only one filesystem in the Rook cluster, there should be no issues and the mount should succeed. If you have more than one filesystem, inconsistent results may arise and the filesystem mounted may not be the one you specified.
If the issue is still not resolved from the steps above, please come chat with us on the **#general** channel of our [Rook Slack](https://rook-slackin.herokuapp.com/).
We want to help you get your storage working and learn from those lessons to prevent users in the future from seeing the same issue.
Create a [rook-tools pod](./toolbox.md) to investigate the current state of CEPH. Here is an example of what one might see. In this case the `ceph status` command would just hang so a CTRL-C needed to be send. The `rookctl status` command is able to give a good amount of detail. In some cases the rook-api pod needs to be restarted for `rookctl` to be able to gather information. If the rook-api is restarted, then the rook-tools pod should be restarted also.
Another indication is when one or more of the MON pods restart frequently. Note the 'mon107' that has only been up for 16 minutes in the following output.
What is happening here is that the MON pods are restarting and one or more of the CEPH daemons are not getting configured with the proper cluster information. This is commonly the result of not specifying a value for `dataDirHostPath` in your Cluster CRD.
The `dataDirHostPath` setting specifies a path on the local host for the CEPH daemons to store configuration and data. Setting this to a path like `/var/lib/rook`, reapplying your Cluster CRD and restarting all the CEPH daemons (MON, MGR, OSD, RGW) should solve this problem. After the CEPH daemons have been restarted, it is advisable to restart the rook-api and [rook-tool Pods](./toolbox.md).
When the operator is starting a cluster, the operator will start one mon at a time and check that they are healthy before continuing to bring up all three mons.
If the first mon is not detected healthy, the operator will continue to check until it is healthy. There are two likely causes for the mon health not being detected:
- The operator pod does not have network connectivity to the mon pod
Likely you will see an error similar to the following that the operator is timing out when connecting to the mon. The last command is `ceph mon_status`,
If the mon pod is failing as in this example, you will need to look at the mon pod status or logs to determine the cause. If the pod is in a crash loop backoff state,
This is a common problem reinitializing the Rook cluster when the local directory used for persistence has **not** been purged.
This directory is the `dataDirHostPath` setting in the cluster CRD and is typically set to `/var/lib/rook`.
To fix the issue you will need to delete all components of Rook and then delete the contents of `/var/lib/rook` (or the directory specified by `dataDirHostPath`) on each of the hosts in the cluster.
Then when the cluster CRD is applied to start a new cluster, the rook-operator should start all the pods as expected.
# OSD pods are failing to start
## Symptoms
* OSD pods are failing to start
* You have started a cluster after tearing down another cluster
## Investigation
When an OSD starts, the device or directory will be configured for consumption. If there is an error with the configuration, the pod will crash and you will see the CrashLoopBackoff
status for the pod. Look in the osd pod logs for an indication of the failure.
```
$ kubectl -n rook logs rook-ceph-osd-fl8fs
...
```
One common case for failure is that you have re-deployed a test cluster and some state may remain from a previous deployment.
If your cluster is larger than a few nodes, you may get lucky enough that the monitors were able to start and form quorum. However, now the OSDs pods may fail to start due to the
old state. Looking at the OSD pod logs you will see an error about the file already existing.
If the error is from the file that already exists, this is a common problem reinitializing the Rook cluster when the local directory used for persistence has **not** been purged.
This directory is the `dataDirHostPath` setting in the cluster CRD and is typically set to `/var/lib/rook`.
To fix the issue you will need to delete all components of Rook and then delete the contents of `/var/lib/rook` (or the directory specified by `dataDirHostPath`) on each of the hosts in the cluster.
Then when the cluster CRD is applied to start a new cluster, the rook-operator should start all the pods as expected.
* After issuing a `reboot` command, node never returned online
* Only a power cycle helps
## Solution
The node needs to be [drained](https://kubernetes.io/docs/tasks/administer-cluster/safely-drain-node/) before reboot. After the successful drain, the node can be rebooted as usual.
Because `kubectl drain` command automatically marks the node as unschedulable (`kubectl cordon` effect), the node needs to be uncordoned once it's back online.
