Jump to contentJump to page navigation: previous page [access key p]/next page [access key n]
documentation.suse.com / SUSE Enterprise Storage 7.1マニュアル / Deploying and Administering SUSE Enterprise Storage with Rook / Troubleshooting Ceph on SUSE CaaS Platform / Common issues
Applies to SUSE Enterprise Storage 7.1

16 Common issues

16.1 Ceph common issues

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 on the configuration of Rook. If the majority of the symptoms are seen, then there is a fair chance that you are experiencing that problem.

16.1.1 Troubleshooting techniques

There are two main categories of information you will need to investigate issues in the cluster:

  1. Kubernetes status and logs.

  2. Ceph cluster status.

16.1.1.1 Running Ceph tools

After you verify the basic health of the running pods, next you will want to run Ceph tools for status of the storage components. There are two ways to run the Ceph tools, either in the Rook toolbox or inside other Rook pods that are already running.

  • Logs on a specific node to find why a PVC is failing to mount: Rook agent errors around the attach and detach:

    kubectl@adm > kubectl logs -n rook-ceph rook-ceph-agent-pod
  • See the Section 12.1.3, “Collecting logs” for a script that will help you gather the logs.

  • Other artifacts:

    • The monitors that are expected to be in quorum:

      kubectl@adm > kubectl -n <cluster-namespace> get configmap rook-ceph-mon-endpoints -o yaml | grep data
16.1.1.1.1 Using tools in the Rook toolbox

The rook-ceph-tools pod provides a simple environment to run Ceph tools. Once the pod is up and running, connect to the pod to execute Ceph commands to evaluate that current state of the cluster.

kubectl@adm > kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l "app=rook-ceph-tools" -o jsonpath='{.items[0].metadata.name}') bash
16.1.1.1.2 Ceph commands

Here are some common commands to troubleshoot a Ceph cluster:

  • ceph status

  • ceph osd status

  • ceph osd df

  • ceph osd utilization

  • ceph osd pool stats

  • ceph osd tree

  • ceph pg stat

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.

16.1.2 Cluster failing to service requests

16.1.2.1 Identifying symptoms

  • Execution of the Ceph command hangs.

  • PersistentVolumes are not being created.

  • Large amount of slow requests are blocking.

  • Large amount of stuck requests are blocking.

  • One or more MONs are restarting periodically.

16.1.2.2 Investigating the current state of Ceph

Create a rook-ceph-tools pod to investigate the current state of Ceph. The following is an example of the output. In this case, the ceph status command would just hang and the process would need to be killed.

kubectl@adm > kubectl -n rook-ceph exec -it $(kubectl -n rook-ceph get pod -l "app=rook-ceph-tools" -o jsonpath='{.items[0].metadata.name}') bash
cephuser@adm > ceph status
^CCluster connection interrupted or timed out

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.

kubectl@adm > kubectl -n rook-ceph get all -o wide --show-all
  NAME                                 READY     STATUS    RESTARTS   AGE       IP               NODE
  po/rook-ceph-mgr0-2487684371-gzlbq   1/1       Running   0          17h       192.168.224.46   k8-host-0402
  po/rook-ceph-mon107-p74rj            1/1       Running   0          16m       192.168.224.28   k8-host-0402
  rook-ceph-mon1-56fgm                 1/1       Running   0          2d        192.168.91.135   k8-host-0404
  rook-ceph-mon2-rlxcd                 1/1       Running   0          2d        192.168.123.33   k8-host-0403
  rook-ceph-osd-bg2vj                  1/1       Running   0          2d        192.168.91.177   k8-host-0404
  rook-ceph-osd-mwxdm                  1/1       Running   0          2d        192.168.123.31   k8-host-0403

16.1.2.3 Identifying the solution

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 Object Gateway daemons have been restarted, it is advisable to restart the rook-tools pod.

16.1.3 Monitors are the only PODs running

16.1.3.1 Identifying symptoms

  • Rook operator is running.

  • Either a single mon starts or the MONs skip letters, specifically named a, d, and f.

  • No MGR, OSD, or other daemons are created.

16.1.3.2 Investigating MON health

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. If the first MON fails to start, a second and then a third MON may attempt to start. However, they will never form a quorum, and orchestration will be blocked from proceeding.

The likely causes for the MON health not being detected:

  • The operator pod does not have network connectivity to the MON pod.

  • The MON pod is failing to start.

  • One or more MON pods are in running state, but are not able to form a quorum.

16.1.3.2.1 Failing to connect to the MON

Firstly, look at the logs of the operator to confirm if it is able to connect to the MONs.

kubectl@adm > kubectl -n rook-ceph logs -l app=rook-ceph-operator

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, followed by a timeout message five minutes later.

  2018-01-21 21:47:32.375833 I | exec: Running command: ceph mon_status --cluster=rook --conf=/var/lib/rook/rook-ceph/rook.config --keyring=/var/lib/rook/rook-ceph/client.admin.keyring --format json --out-file /tmp/442263890
  2018-01-21 21:52:35.370533 I | exec: 2018-01-21 21:52:35.071462 7f96a3b82700  0 monclient(hunting): authenticate timed out after 300
  2018-01-21 21:52:35.071462 7f96a3b82700  0 monclient(hunting): authenticate timed out after 300
  2018-01-21 21:52:35.071524 7f96a3b82700  0 librados: client.admin authentication error (110) Connection timed out
  2018-01-21 21:52:35.071524 7f96a3b82700  0 librados: client.admin authentication error (110) Connection timed out
  [errno 110] error connecting to the cluster

The error would appear to be an authentication error, but it is misleading. The real issue is a timeout.

16.1.3.2.2 Identifying the solution

If you see the timeout in the operator log, verify if the MON pod is running (see the next section). If the MON pod is running, check the network connectivity between the operator pod and the MON pod. A common issue is that the CNI is not configured correctly.

16.1.3.2.3 Failing MON pod

We need to verify if the MON pod started successfully.

kubectl@adm > kubectl -n rook-ceph get pod -l app=rook-ceph-mon
NAME                                READY     STATUS               RESTARTS   AGE
rook-ceph-mon-a-69fb9c78cd-58szd    1/1       CrashLoopBackOff     2          47s

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, you should see the reason by describing the pod.

The pod shows a termination status that the keyring does not match the existing keyring.

kubectl@adm > kubectl -n rook-ceph describe pod -l mon=rook-ceph-mon0
[...]
Last State:    Terminated
Reason:    Error
Message:    The keyring does not match the existing keyring in /var/lib/rook/rook-ceph-mon0/data/keyring.
You may need to delete the contents of dataDirHostPath on the host from a previous deployment.
[...]

See the solution in the next section regarding cleaning up the dataDirHostPath on the nodes.

If you see the three mons running with the names a, d, and f, they likely did not form quorum even though they are running.

NAME                               READY   STATUS    RESTARTS   AGE
rook-ceph-mon-a-7d9fd97d9b-cdq7g   1/1     Running   0          10m
rook-ceph-mon-d-77df8454bd-r5jwr   1/1     Running   0          9m2s
rook-ceph-mon-f-58b4f8d9c7-89lgs   1/1     Running   0          7m38s
16.1.3.2.4 Identifying the solution

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.

Important
Important

Deleting the dataDirHostPath folder is destructive to the storage. Only delete the folder if you are trying to permanently purge the Rook cluster.

16.1.4 PVCs stay in pending state

16.1.4.1 Identifying symptoms

When you create a PVC based on a Rook storage class, it stays pending indefinitely.

For the Wordpress example, you might see two PVCs in the pending state.

kubectl@adm > kubectl get pvc
NAME             STATUS    VOLUME   CAPACITY   ACCESS MODES   STORAGECLASS      AGE
mysql-pv-claim   Pending                                      rook-ceph-block   8s
wp-pv-claim      Pending                                      rook-ceph-block   16s

16.1.4.2 Investigating common causes

There are two common causes for the PVCs staying in the pending state:

  1. There are no OSDs in the cluster.

  2. The CSI provisioner pod is not running or is not responding to the request to provision the storage.

16.1.4.2.1 Confirming if there are OSDs

To confirm if you have OSDs in your cluster, connect to the Rook Toolbox and run the ceph status command. You should see that you have at least one OSD up and in. The minimum number of OSDs required depends on the replicated.size setting in the pool created for the storage class. In a test cluster, only one OSD is required (see storageclass-test.yaml). In the production storage class example (storageclass.yaml), three OSDs would be required.

cephuser@adm > ceph status
  cluster:
  id:     a0452c76-30d9-4c1a-a948-5d8405f19a7c
  health: HEALTH_OK

  services:
  mon: 3 daemons, quorum a,b,c (age 11m)
  mgr: a(active, since 10m)
  osd: 1 osds: 1 up (since 46s), 1 in (since 109m)
16.1.4.2.2 Preparing OSD logs

If you do not see the expected number of OSDs, investigate why they were not created. On each node where Rook looks for OSDs to configure, you will see an osd prepare pod.

kubectl@adm > kubectl -n rook-ceph get pod -l app=rook-ceph-osd-prepare
NAME                                 ...  READY   STATUS      RESTARTS   AGE
rook-ceph-osd-prepare-minikube-9twvk   0/2     Completed   0          30m

See the section on Section 16.1.6, “OSD pods are not created on my devices” to investigate the logs.

16.1.4.2.3 Checking CSI driver

The CSI driver may not be responding to the requests. Look in the logs of the CSI provisioner pod to see if there are any errors during the provisioning.

There are two provisioner pods:

kubectl@adm > kubectl -n rook-ceph get pod -l app=csi-rbdplugin-provisioner

Get the logs of each of the pods. One of them should be the leader and be responding to requests.

kubectl@adm > kubectl -n rook-ceph logs csi-cephfsplugin-provisioner-d77bb49c6-q9hwq csi-provisioner
16.1.4.2.4 Restarting the operator

Lastly, if you have OSDs up and in, the next step is to confirm the operator is responding to the requests. Look in the operator pod logs around the time when the PVC was created to confirm if the request is being raised. If the operator does not show requests to provision the block image, the operator may be stuck on some other operation. In this case, restart the operator pod to get things going again.

16.1.4.3 Identifying the solution

If the OSD prepare logs did not give you enough clues about why the OSDs were not being created, review your cluster.yaml configuration. The common mistakes include:

  • If useAllDevices: true, Rook expects to find local devices attached to the nodes. If no devices are found, no OSDs will be created.

  • If useAllDevices: false, OSDs will only be created if deviceFilter is specified.

  • Only local devices attached to the nodes will be configurable by Rook. In other words, the devices must show up under /dev.

    • The devices must not have any partitions or file systems on them. Rook will only configure raw devices. Partitions are not yet supported.

16.1.5 OSD pods are failing to start

16.1.5.1 Identifying symptoms

  • OSD pods are failing to start.

  • You have started a cluster after tearing down another cluster.

16.1.5.2 Investigating configuration errors

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@adm > kubectl -n rook-ceph 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 a 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.

kubectl -n rook-ceph logs rook-ceph-osd-fl8fs
[...]
2017-10-31 20:13:11.187106 I | mkfs-osd0: 2017-10-31 20:13:11.186992 7f0059d62e00 -1 bluestore(/var/lib/rook/osd0) _read_fsid unparsable uuid
2017-10-31 20:13:11.187208 I | mkfs-osd0: 2017-10-31 20:13:11.187026 7f0059d62e00 -1 bluestore(/var/lib/rook/osd0) _setup_block_symlink_or_file failed to create block symlink to /dev/disk/by-partuuid/651153ba-2dfc-4231-ba06-94759e5ba273: (17) File exists
2017-10-31 20:13:11.187233 I | mkfs-osd0: 2017-10-31 20:13:11.187038 7f0059d62e00 -1 bluestore(/var/lib/rook/osd0) mkfs failed, (17) File exists
2017-10-31 20:13:11.187254 I | mkfs-osd0: 2017-10-31 20:13:11.187042 7f0059d62e00 -1 OSD::mkfs: ObjectStore::mkfs failed with error (17) File exists
2017-10-31 20:13:11.187275 I | mkfs-osd0: 2017-10-31 20:13:11.187121 7f0059d62e00 -1  ** ERROR: error creating empty object store in /var/lib/rook/osd0: (17) File exists

16.1.5.3 Solution

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.

16.1.6 OSD pods are not created on my devices

16.1.6.1 Identifying symptoms

  • No OSD pods are started in the cluster.

  • Devices are not configured with OSDs even though specified in the cluster CRD.

  • One OSD pod is started on each node instead of multiple pods for each device.

16.1.6.2 Investigating

First, ensure that you have specified the devices correctly in the CRD. The cluster CRD has several ways to specify the devices that are to be consumed by the Rook storage:

  • useAllDevices: true: Rook will consume all devices it determines to be available.

  • deviceFilter: Consume all devices that match this regular expression.

  • devices: Explicit list of device names on each node to consume.

Second, if Rook determines that a device is not available (has existing partitions or a formatted file system), Rook will skip consuming the devices. If Rook is not starting OSDs on the devices you expect, Rook may have skipped it for this reason. To see if a device was skipped, view the OSD preparation log on the node where the device was skipped. Note that it is completely normal and expected for OSD prepare pod to be in the completed state. After the job is complete, Rook leaves the pod around in case the logs need to be investigated.

Get the prepare pods in the cluster:

kubectl@adm > kubectl -n rook-ceph get pod -l app=rook-ceph-osd-prepare
NAME                                   READY     STATUS      RESTARTS   AGE
rook-ceph-osd-prepare-node1-fvmrp      0/1       Completed   0          18m
rook-ceph-osd-prepare-node2-w9xv9      0/1       Completed   0          22m
rook-ceph-osd-prepare-node3-7rgnv      0/1       Completed   0          22m

View the logs for the node of interest in the "provision" container:

kubectl@adm > kubectl -n rook-ceph logs rook-ceph-osd-prepare-node1-fvmrp provision

Here are some key lines to look for in the log. A device will be skipped if Rook sees it has partitions or a file system:

2019-05-30 19:02:57.353171 W | cephosd: skipping device sda that is in use
2019-05-30 19:02:57.452168 W | skipping device "sdb5": ["Used by ceph-disk"]

Other messages about a disk being unusable by Ceph include:

Insufficient space (<5GB) on vgs
Insufficient space (<5GB)
LVM detected
Has BlueStore device label
locked
read-only

A device is going to be configured:

2019-05-30 19:02:57.535598 I | cephosd: device sdc to be configured by ceph-volume

For each device configured, you will see a report in the log:

2019-05-30 19:02:59.844642 I |   Type            Path                                                    LV Size         % of device
2019-05-30 19:02:59.844651 I | ----------------------------------------------------------------------------------------------------
2019-05-30 19:02:59.844677 I |   [data]          /dev/sdc                                                7.00 GB         100%

16.1.6.3 Solution

Either update the CR with the correct settings, or clean the partitions or file system from your devices.

After the settings are updated or the devices are cleaned, trigger the operator to analyze the devices again by restarting the operator. Each time the operator starts, it will ensure all the desired devices are configured. The operator does automatically deploy OSDs in most scenarios, but an operator restart will cover any scenarios that the operator does not detect automatically.

Restart the operator to ensure devices are configured. A new pod will automatically be started when the current operator pod is deleted.

kubectl@adm > kubectl -n rook-ceph delete pod -l app=rook-ceph-operator

16.1.7 Rook agent modprobe exec format error

16.1.7.1 Identifying symptoms

  • PersistentVolumes from Ceph fail or timeout to mount.

  • Rook Agent logs contain modinfo: ERROR: could not get modinfo from 'rbd': Exec format error lines.

16.1.7.2 Solution

If it is feasible to upgrade your kernel, you should upgrade to 4.x, even better is 4.7 or above, due to a feature for CephFS added to the kernel.

If you are unable to upgrade the kernel, you need to go to each host that will consume storage and run:

modprobe rbd

This command inserts the rbd module into the kernel.

To persist this fix, you need to add the rbd kernel module to either /etc/modprobe.d/ or /etc/modules-load.d/. For both paths create a file called rbd.conf with the following content:

rbd

Now when a host is restarted, the module should be loaded automatically.

16.1.8 Using multiple shared file systems (CephFS) is attempted on a kernel version older than 4.7

16.1.8.1 Identifying symptoms

  • More than one shared file system (CephFS) has been created in the cluster.

  • A pod attempts to mount any other shared file system besides the first one that was created.

  • The pod incorrectly gets the first file system mounted instead of the intended file system.

16.1.8.2 Solution

The only solution to this problem is to upgrade your kernel to 4.7 or higher. This is due to a mount flag added in kernel version 4.7, which allows choosing the file system by name.

16.1.9 Activating log to file for a particular Ceph daemon

They are cases where looking at Kubernetes logs is not enough for various reasons, but just to name a few:

  • Not everyone is familiar for Kubernetes logging and expects to find logs in traditional directories.

  • Logs get eaten (buffer limit from the log engine) and thus not requestable from Kubernetes.

So for each daemon, dataDirHostPath is used to store logs, if logging is activated. Rook will bind-mount dataDirHostPath for every pod. As of Ceph Nautilus 14.2.1, it is possible to enable logging for a particular daemon on the fly. Let us say you want to enable logging for mon.a, but only for this daemon. Using the toolbox or from inside the operator run:

cephuser@adm > ceph config daemon mon.a log_to_file true

This will activate logging on the file system, you will be able to find logs in dataDirHostPath/$NAMESPACE/log, so typically this would mean /var/lib/rook/rook-ceph/log. You do not need to restart the pod, the effect will be immediate.

To disable the logging on file, simply set log_to_file to false.

For Ceph Luminous and Mimic releases, mon_cluster_log_file and cluster_log_file can be set to /var/log/ceph/XXXX in the config override ConfigMap to enable logging.

16.1.10 A worker node using RBD devices hangs up

16.1.10.1 Identifying symptoms

  • There is no progress on I/O from/to one of RBD devices (/dev/rbd* or /dev/nbd*).

  • After that, the whole worker node hangs up.

16.1.10.2 Investigating

This happens when the following conditions are satisfied.

  • The problematic RBD device and the corresponding OSDs are co-located.

  • There is an XFS file system on top of this device.

In addition, when this problem happens, you can see the following messages in dmesg.

dmesg
...
[51717.039319] INFO: task kworker/2:1:5938 blocked for more than 120 seconds.
[51717.039361]       Not tainted 4.15.0-72-generic #81-Ubuntu
[51717.039388] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
...

This is the so-called hung_task problem and means that there is a deadlock in the kernel.

16.1.10.3 Solution

You can bypass this problem by using ext4 or any other file systems rather than XFS. The file system type can be specified with csi.storage.k8s.io/fstype in StorageClass resource.

16.1.11 Too few PGs per OSD warning is shown

16.1.11.1 Identifying symptoms

  • ceph status shows too few PGs per OSD warning as follows.

cephuser@adm > ceph status
cluster:
id:     fd06d7c3-5c5c-45ca-bdea-1cf26b783065
health: HEALTH_WARN
too few PGs per OSD (16 < min 30)

16.1.11.2 Solution

See Chapter 5, Troubleshooting placement groups (PGs) for more information.

16.1.12 LVM metadata can be corrupted with OSD on LV-backed PVC

16.1.12.1 Identifying symptoms

There is a critical flaw in OSD on LV-backed PVC. LVM metadata can be corrupted if both the host and OSD container modify it simultaneously. For example, the administrator might modify it on the host, while the OSD initialization process in a container could modify it too. In addition, if lvmetad is running, the possibility of occurrence gets higher. In this case, the change of LVM metadata in OSD container is not reflected to LVM metadata cache in host for a while.

If you still decide to configure an OSD on LVM, keep the following in mind to reduce the probability of this issue.

16.1.12.2 Solution

  • Disable lvmetad.

  • Avoid configuration of LVs from the host. In addition, do not touch the VGs and physical volumes that back these LVs.

  • Avoid incrementing the count field of storageClassDeviceSets and create a new LV that backs a OSD simultaneously.

You can know whether the above-mentioned tag exists tag by running

#  lvs -o lv_name,lv_tags

If the lv_tag field is empty in an LV corresponding to the OSD lv_tags, this OSD encountered the problem. In this case, retire this OSD or replace with other new OSD before restarting.