Troubleshooting MySQL for Pivotal Platform

Page last updated:

This topic provides operators with basic instructions for troubleshooting on-demand MySQL for Pivotal Cloud Foundry (PCF). For information about temporary MySQL for PCF service interruptions, see Service Interruptions.

Troubleshoot Errors

This section provides information on how to troubleshoot specific errors or error messages.

Common Services Errors

The following errors occur in multiple services:


Failed Installation

Symptom MySQL for PCF fails to install.
Cause Reasons for a failed installation include:
  • Certificate issues: The on-demand broker (ODB) requires valid certificates.
  • Deploy fails. This could be due to a variety of reasons.
  • Networking problems:
    • Cloud Foundry cannot reach the MySQL for PCF broker
    • Cloud Foundry cannot reach the service instances
    • The service network cannot access the BOSH director
  • The Register broker errand fails.
  • The smoke test errand fails.
  • Resource sizing issues: These occur when the resource sizes selected for a given plan are less than MySQL for PCF requires to function.
  • Other service-specific issues.
Solution To troubleshoot:
  • Certificate issues: Ensure that your certificates are valid and generate new ones if necessary. To generate new certificates, contact Support.
  • Deploy fails: View the logs using Ops Manager to determine why the deploy is failing.
  • Networking problems: For how to troubleshoot, see Networking problems.
  • Register broker errand fails: For how to troubleshoot, see Register broker errand.
  • Resource sizing issues: Check your resource configuration in Ops Manager and ensure that the configuration matches that recommended by the service.


Cannot Create or Delete Service Instances

Symptom If developers report errors such as: 
Instance provisioning failed: There was a problem completing your request. Please contact your operations team providing the following information: service: redis-acceptance, service-instance-guid: ae9e232c-0bd5-4684-af27-1b08b0c70089, broker-request-id: 63da3a35-24aa-4183-aec6-db8294506bac, task-id: 442, operation: create
Cause Reasons include:
  • Problems with the deployment manifest
  • Authentication errors
  • Network errors
  • Quota errors
Solution To troubleshoot:

  1. If the BOSH error shows a problem with the deployment manifest, open the manifest in a text editor to inspect it.

  2. To continue troubleshooting, Log in to BOSH and target the MySQL for PCF instance using the instructions on parsing a Cloud Foundry error message.

  3. Retrieve the BOSH task ID from the error message and run the following command: 

    bosh task TASK-ID

  4. If you need more information, access the broker logs and use the broker-request-id from the error message above to search the logs for more information. Check for:


Broker Request Timeouts

Symptom If developers report errors such as: 
Server error, status code: 504, error code: 10001, message: The request to the service broker timed out: https://BROKER-URL/v2/service_instances/e34046d3-2379-40d0-a318-d54fc7a5b13f/service_bindings/aa635a3b-ef6d-41c3-a23f-55752f3f651b
Cause Cloud Foundry might not be connected to the service broker, or there might be a large number of queued tasks.
Solution To troubleshoot:
  1. Confirm that Cloud Foundry (CF) is connected to the service broker.
  2. Check the BOSH queue size:
    1. Log in to BOSH as an admin.
    2. Run 
      bosh tasks
    If there are a large number of queued tasks, the system may be under too much load. BOSH is configured with two workers and one status worker, which might not be sufficient resources for the level of load.
  3. If the task queue is long, advise app developers to try again once the system is under less load.


Instance Does Not Exist

Symptom If developers report errors such as: 
Server error, status code: 502, error code: 10001, message: Service broker error: instance does not exist`
Cause The instance might have been deleted.
Solution To troubleshoot:

  1. Confirm that the MySQL for PCF instance exists in BOSH and obtain the GUID CF by running: 

    cf service MY-INSTANCE --guid

  2. Using the GUID obtained above, run: 

    bosh -d service-instance_GUID vms

If the BOSH deployment is not found, it has been deleted from BOSH. Contact Support for further assistance.


Cannot Bind to or Unbind from Service Instances

Symptom If developers report errors such as: 
Server error, status code: 502, error code: 10001, message: Service broker error: There was a problem completing your request. Please contact your operations team providing the following information: service: example-service, service-instance-guid: 8d69de6c-88c6-4283-b8bc-1c46103714e2, broker-request-id: 15f4f87e-200a-4b1a-b76c-1c4b6597c2e1, operation: bind
Cause This might be due to authentication or network errors.
Solution To find out the exact issue with the binding process:

  1. Access the service broker logs.

  2. Search the logs for the broker-request-id string listed in the error message above.

  3. Check for:

  4. Contact Support for further assistance if you are unable to resolve the problem.


Cannot Connect to a Service Instance

Symptom Developers report that their app cannot use service instances that they have successfully created and bound.
Cause The error might originate from the service or be network related.
Solution To solve this issue, ask the user to send application logs that show the connection error. If the error originates from the service, then follow MySQL for PCF-specific instructions. If the issue appears to be network-related, then:

  1. Check that application security groups are configured correctly. Access should be configured for the service network that the tile is deployed to.

  2. Ensure that the network the PAS tile is deployed to has network access to the service network. You can find the network definition for this service network in the BOSH Director tile.

  3. In Ops Manager go into the service tile and see the service network that is configured in the networks tab.

  4. In Ops Manager go into the PAS tile and see the network it is assigned to. Make sure that these networks can access each other.

Note: Service instances can also become temporarily inaccessible during upgrades and VM or network failures. See Service Interruptions for more information.


Upgrade All Service Instances Errand Fails

Symptom The upgrade-all-service-instances errand fails.
Cause There might be a problem with a particular instance.
Solution To troubleshoot:
  1. Look at the errand output in the Ops Manager log.
  2. If an instance has failed to upgrade, debug and fix it before running the errand again to prevent any failure issues from spreading to other on-demand instances.
  3. After the Ops Manager log no longer lists the deployment as failing, re-run the errand to upgrade the rest of the instances.


Missing Logs and Metrics

Symptom No logs are being emitted by the on-demand broker.
Cause Syslog might not be configured correctly, or you might have network access issues.
Solution To troubleshoot:

  1. Ensure you have configured syslog for the tile.

  2. Check that your syslog forwarding address is correct in Ops Manager.

  3. Ensure that you have network connectivity between the networks that the tile is using and the syslog destination. If the destination is external, you need to use the public ip VM extension feature available in your Ops Manager tile configuration settings.

  4. Verify that Loggregator is emitting metrics:

    1. Install the cf log-stream plugin. For instructions, see the Log Stream CLI Plugin GitHub repository.

    2. Find the GUID for your service instance by running: 

      cf service SERVICE-INSTANCE --guid

    3. Find logs from your service instance by running: 

      cf log-stream | grep "SERVICE-GUID"
    4. If no metrics appear within five minutes, verify that the broker network has access to the Loggregator system on all required ports.
  5. If you are unable to resolve the issue, contact Support.

Leader-Follower Service Instance Errors

This section provides solutions for the following errands:


Unable to Determine Leader and Follower

Symptom This problem happens when the configure-leader-follower errand fails because it cannot determine the VM roles.

The configure-leader-follower errand exits with 1 and the errand logs contain the following: 
$ Unable to determine leader and follower based on transaction history.
Cause Something has happened to the instances, such as a failure or manual intervention. As a result, there is not enough information available to determine the correct state and topology without operator intervention to resolve the issue.
Solution Use the inspect errand to determine which instance should be the leader. Then, using the orchestration errands and backup/restore, you can put the service instance into a safe topology, and then rerun the configure-leader-follower errand. This is shown in the example below.

This example shows one outcome that the inspect errand can return:
  1. Use the inspect errand to retrieve relevant information about the two VMs: 
    $ bosh -e my-env -d my-dep run-errand inspect
      [...]
      Instance   mysql/4ecad54b-0704-47eb-8eef-eb228cab9724
      Exit Code  0
      Stdout     -
      Stderr   2017/12/11 18:25:54 Started executing command: inspect
               2017/12/11 18:25:54 Started GET https://127.0.0.1:8443/status
               2017/12/11 18:25:54
               Has Data: false
               Read Only: true
               GTID Executed: 1d774323-de9e-11e7-be01-42010a001014:1-25
               Replication Configured: false
      Instance   mysql/e0b94ade-0114-4d49-a929-ce1616d8beda
      Exit Code  0
      Stdout     -
      Stderr   2017/12/11 18:25:54 Started executing command: inspect
               2017/12/11 18:25:54
               Started GET https://127.0.0.1:8443/status
               2017/12/11 18:25:54
               Has Data: true
               Read Only: true
               GTID Executed: 1d774323-de9e-11e7-be01-42010a001014:1-25
               Replication Configured: true
      2 errand(s)
      Succeeded
    In the above scenario, the first instance is missing data but does not have replication configured. The second instance has data, and also has replication configured. The instructions below resolve this by copying data to the first instance, and resuming replication.
  2. Take a backup of the second instance using the Create a MySQL for Pivotal Cloud Foundry Logical Backup steps.
  3. Restore the backup artifact to the first instance using the Restore from a MySQL for Pivotal Cloud Foundry Logical Backup steps.
    At this point, the instances have equivalent data.
  4. Run the configure-leader-follower errand to reconfigure replication: 
    bosh -e ENVIRONMENT -d DEPLOYMENT \
          run-errand configure-leader-follower \
          --instance=mysql/GUID-OF-LEADER
    For example: 
    $ bosh -e my-env -d my-dep \
          run-errand configure-leader-follower \
          --instance=mysql/4ecad54b-0704-47eb-8eef-eb228cab9724


Both Leader and Follower Instances Are Writable

Symptom This problem happens when theconfigure-leader-follower errand fails because both VMs are writable and the VMs might hold differing data.

The configure–leader-follower errand exits with 1 and the errand logs contain the following: 
$ Both mysql instances are writable. Please ensure no divergent data and set one instance to read-only mode.
Cause MySQL for Pivotal Cloud Foundry tries to ensure that there is only one writable instance of the leader-follower pair at any given time. However, in certain situations, such as network partitions, or manual intervention outside of the provided bosh errands, it is possible for both instances to be writable.

The service instances remain in this state until an operator resolves the issue to ensure that the correct instance is promoted and reduce the potential for data divergence.
Solution
  1. Use the inspect errand to retrieve the GTID Executed set for each VM: 
    $ bosh -e my-env -d my-dep run-errand inspect
        [...]
        Instance   mysql/4ecad54b-0704-47eb-8eef-eb228cab9724
        Exit Code  0
        Stdout     -
        Stderr     2017/12/11 18:25:54 Started executing command: inspect
                 2017/12/11 18:25:54 Started GET https:127.0.0.1:8443/status
                 2017/12/11 18:25:54
                 Has Data: true
                 Read Only: false
                 GTID Executed: 1d774323-de9e-11e7-be01-42010a001014:1-23
                 Replication Configured: false

    
        Instance   mysql/e0b94ade-0114-4d49-a929-ce1616d8beda
        Exit Code  0
        Stdout     -
        Stderr     2017/12/11 18:25:54 Started executing command: inspect
                 2017/12/11 18:25:54 Started GET https:127.0.0.1:8443/status
                 2017/12/11 18:25:54
                 Has Data: true
                 Read Only: false
                 GTID Executed: 1d774323-de9e-11e7-be01-42010a001014:1-25
                 Replication Configured: false

    
        2 errand(s)
    
        Succeeded
    If the GTID Executed sets for both instances are the same, continue to Step 2. If they are different, continue to Step 4.
  2. Look at the value of GTID Executed for both instances.
    • If the range after the GUID is equivalent, either instance can be made read-only, as described in Step 3.
    • If one instance has a range that is a subset of the other, the instance with the subset should must be made read-only, as described in Step 3.
  3. Based on the information you gathered in the step above, run the make-read-only errand to make the appropriate instance read-only: 
    bosh -e ENVIRONMENT -d DEPLOYMENT \
            run-errand make-read-only \
            --instance=mysql/MYSQL-SUBSET-INSTANCE
    For example: 
    $ bosh -e my-env -d my-dep \
              run-errand make-read-only \
              --instance=mysql/e0b94ade-0114-4d49-a929-ce1616d8beda
            [...]
            succeeded
  4. If the GTID Executed sets are neither equivalent nor subsets, data has diverged and you must determine what data has diverged as part of the procedure below:
    1. Use the make-read-only errand to set both instances to read-only to prevent further data divergence. 
      bosh -e ENVIRONMENT -d DEPLOYMENT \
                run-errand make-read-only \
                --instance=mysql/MYSQL-INSTANCE
      For example: 
      $ bosh -e my-env -d my-dep \
                  run-errand make-read-only \
                  --instance=mysql/e0b94ade-0114-4d49-a929-ce1616d8beda
                  [...]
                  succeeded
    2. Take a backup of both instances using the Create a MySQL for Pivotal Cloud Foundry Logical Backup steps.
    3. Manually inspect the data on each instance to determine the discrepancies and put the data on the instance that is further ahead—this instance has the higher GTID Executed set, and will be the new leader.
    4. Migrate all appropriate data to the new leader instance.
    5. After putting all data on the leader, ssh onto the follower: 
      bosh -e ENVIRONMENT -d DEPLOYMENT ssh mysql/GUID-OF-FOLLOWER
      For example: 
      $ bosh -e my-env -d my-dep ssh mysql/e0b94ade-0114-4d49-a929-ce1616d8beda
    6. Become root with the command sudo su.
    7. Stop the mysql process with the command monit stop mysql.
    8. Delete the data directory of the follower with the command rm -rf /var/vcap/store/mysql.
    9. Start the mysql process with the command monit start mysql.
    10. Use the configure-leader-follower errand to copy the leader data to the follower and resume replication: 
      bosh -e ENVIRONMENT -d DEPLOYMENT \
                run-errand configure-leader-follower \
                --instance=mysql/GUID-OF-LEADER
      For example: 
      $ bosh -e my-env -d my-dep \
                run-errand configure-leader-follower \
                --instance=mysql/4ecad54b-0704-47eb-8eef-eb228cab9724


Both Leader and Follower Instances Are Read-Only

Symptom Developers report that apps cannot write to the database. In a leader-follower topology, the leader VM is writable and the follower VM is read-only. However if both VMs are read only, apps cannot write to the database.
Cause This problem happens if the leader VM fails and the BOSH Resurrector is enabled. When the leader is resurrected, it is set as read-only.
Solution
  1. Use the inspect errand to confirm that both VMs are in a read-only state: 
    bosh -e ENVIRONMENT -d DEPLOYMENT run-errand inspect
  2. Examine the output and locate the information about the leader-follower MySQL for Pivotal Cloud Foundry VMs: 
        Instance   mysql/4eexample54b-0704-47eb-8eef-eb2example724
    Exit Code  0
    Stdout     -
    Stderr     2017/12/11 18:25:54 Started executing command: inspect
             2017/12/11 18:25:54 Started GET https:999.0.0.1:8443/status
             2017/12/11 18:25:54
             Has Data: true
             Read Only: true
             GTID Executed: 1d779999-de9e-11e7-be01-42010a009999:1-23
             Replication Configured: true

    
    Instance   mysql/e0exampleade-0114-4d49-a929-cexample8beda
    Exit Code  0
    Stdout     -
    Stderr     2017/12/11 18:25:54 Started executing command: inspect
             2017/12/11 18:25:54 Started GET https:999.0.0.1:8443/status
             2017/12/11 18:25:54
             Has Data: true
             Read Only: true
             GTID Executed: 1d779999-de9e-11e7-be01-42010a009999:1-25
             Replication Configured: false

    
    2 errand(s)
    
    Succeeded
  3. If Read Only is set to true for both VMs, make the leader writable using the following command: 
    bosh -e ENVIRONMENT -d DEPLOYMENT \
          run-errand configure-leader-follower \
          --instance=mysql/GUID-OF-LEADER
    For example, if the second instance above is the leader: 
        $ bosh -e my-env -d my-dep \
      run-errand configure-leader-follower \
      --instance=mysql/e0exampleade-0114-4d49-a929-cexample8beda

Inoperable App and Database Errors

This section provides a solution for the following errors:


Persistent Disk is Full

Symptom Developers report that read, write, and cf CLI operations do not work. Developers cannot upgrade to a larger MySQL for Pivotal Cloud Foundry service plan to free up disk space.

If your persistent disk is full, apps become inoperable. In this state, read, write, and Cloud Foundry Command-Line Interface (cf CLI) operations do not work.
Cause This problem happens if your persistent disk is full. When you use the BOSH CLI to target your deployment, you see that instances are at 100% persistent disk usage.

Available disk space can be increased by deleting log files. After deleting logs, you can then upgrade to a larger MySQL for Pivotal Cloud Foundry service plan.

You can also turn off binary logging before developers do large data uploads or if their databases have a high transaction volume.
Solution To resolve this issue, do one of the following:


Cannot Access Database Table

Symptom When you query an existing table, you see an error similar to the following: 
ERROR 1146 (42S02): Table 'mysql.foobar' doesn't exist
Cause

This error occurs if you created an uppercase table name and then enabled lowercase table names.

You enable lowercase table names either by:

  • Setting the optional enable_lower_case_table_names parameter to true with the cf CLI. For more information about the parameter, see Lowercase Table Names.
  • Selecting Enable Lower Case Table Names in the Mysql Configuration pane of the tile. For more information about this configuration, see Configure MySQL.
Solution

To resolve this issue:

  1. Disable lowercase table names by doing one of the following:
    • Set the optional enable_lower_case_table_names parameter to false with the cf CLI. For instructions, see Set Optional Parameters.
    • Enable lowercase table names in the tile:
      1. Deselect Enable Lower Case Table Names in the Mysql Configuration pane of the tile.
      2. Navigate to the Ops Manager Installation Dashboard, click Review Pending Changes, and then click Apply Changes.
  2. (Optional) If you want to re-enable lowercase table names, re-name your table to lowercase and then enable lowercase table names.

Highly Available Cluster Errors

This section provides solutions for the following errands:


Unresponsive Node in a Highly Available Cluster

Symptom A client connected to a MySQL for Pivotal Cloud Foundry cluster node reports the following error: 
WSREP has not yet prepared this node for application use
Some clients might instead return the following: 
      unknown error
Cause If the client is connected to a MySQL for Pivotal Cloud Foundry cluster node and that node loses connection to the rest of the cluster, the node stops accepting writes. If the connection to this node is made through the proxy, the proxy automatically re-routes further connections to a different node.
Solution A node can become unresponsive for a number of reasons. For solutions, see the following:
  • Network Latency: If network latency causes a node to become unresponsive, the node drops but eventually rejoins. The node automatically rejoins only if one node has left the cluster. Consult your IaaS network settings to reduce your network latency.
  • MySQL Process Failure: If the MySQL process crashes, monit and BOSH should restore the process. If the process is not restored, run the download-logs tool and consult the error logs it generates. For more information, see the download-logs section below.
  • Firewall Rule Change: If your firewall rules change, it might prevent a node from communicating with the rest of the cluster. This causes the node to become unresponsive. In this case, the logs show the node leaving the cluster but do not show network latency errors.

    To confirm that the node is unresponsive because of a firewall rule change, SSH from a responsive node to the unresponsive node. If you cannot connect, the node is unresponsive due to a firewall rule change. Change your firewall rules to enable the unresponsive node to rejoin the cluster.
  • VM Failure: If you cannot SSH into a node and you are not detecting either network latency or firewall issues, your node might be down due to VM failure. To confirm that the node is unresponsive and re-create the VM, see Re-create a Corrupted VM in a Highly Available below.
  • Node Unable to Rejoin: If a detached existing node fails to join the cluster, its sequence_number might be higher than those of the nodes with quorum. A higher sequence_number on the detached node indicates that it has recent changes to the data that the primary component lacks. You can check this by looking at the node’s error log at /var/vcap/sys/log/pxc-mysql/mysql.err.log.

    To restore the cluster, do one the following:
    • If the detached node has a higher sequence number than the primary component, do the procedures in Bootstrapping.
    • If bootstrapping does not restore the cluster, you can manually force the node to rejoin the cluster. This removes all of the unsynchronized data from the detached server node and creates a new copy of the cluster data on the node. For more information, see Force a Node to Rejoin a Highly Available Cluster Manually below.

      Warning: Forcing a node to rejoin the cluster is a destructive procedure. Only do this procedure with the assistance of Pivotal Support.


Many Replication Errors in Logs for Highly Available Clusters

Symptom You see many replication errors in the MySQL logs, like the following: 
160318 9:25:16 [Warning] WSREP: RBR event 1 Query apply warning: 1, 16992456
160318 9:25:16 [Warning] WSREP: Ignoring error for TO isolated action: source: abcd1234-abcd-1234-abcd-1234abcd1234 version: 3 local: 0 state: APPLYING flags: 65 conn_id: 246804 trx_id: -1 seqnos (l: 865022, g: 16992456, s: 16992455, d: 16992455, ts: 2530660989030983)
160318 9:25:16 [ERROR] Slave SQL: Error 'Duplicate column name 'number'' on query. Default database: 'cf_0123456_1234_abcd_1234_abcd1234abcd'. Query: 'ALTER TABLE ...'
Cause This problem happens when there are errors in SQL statements.
Solution For solutions for replication errors in MySQL log files, see the table below.
Additional Error Solution
ALTER TABLE errors Fix the ALTER TABLE error.
This error can occur when an app issues an invalid data definition statement. Other nodes log this problem as a replication error because they fail to replicate the ALTER TABLE.
Increased persistent disk usage or running out of working memory Decode the GRA_*.log files and look for errors. See How to decode Galera GRA log files in the Pivotal Support knowledge base. The GRA log files contain failing DDL statements.
If you see replication errors, but no ALTER TABLE or persistent disk or memory issues, you can ignore the replication errors.

Troubleshoot Components

This section provides guidance on checking for and fixing issues in on-demand service components.

BOSH Problems

Large BOSH Queue

On-demand service brokers add tasks to the BOSH request queue, which can back up and cause delay under heavy loads. An app developer who requests a new MySQL for PCF instance sees create in progress in the Cloud Foundry Command Line Interface (cf CLI) until BOSH processes the queued request.

Ops Manager currently deploys two BOSH workers to process its queue. Future versions of Ops Manager will let users configure the number of BOSH workers.

Configuration

Service Instances in Failing State

The VM or Disk type that you configured in the plan page of the tile in Ops Manager might not be large enough for the MySQL for PCF service instance to start. See tile-specific guidance on resource requirements.

Authentication

UAA Changes

If you have rotated any UAA user credentials then you may see authentication issues in the service broker logs.

To resolve this, redeploy the MySQL for PCF tile in Ops Manager. This provides the broker with the latest configuration.

Note: You must ensure that any changes to UAA credentials are reflected in the Ops Manager credentials tab of the Pivotal Application Service tile.

Networking

Common issues with networking include:

Issue Solution
Latency when connecting to the MySQL for PCF service instance to create or delete a binding. Try again or improve network performance.
Firewall rules are blocking connections from the MySQL for PCF service broker to the service instance. Open the MySQL for PCF tile in Ops Manager and check the two networks configured in the Networks pane. Ensure that these networks allow access to each other.
Firewall rules are blocking connections from the service network to the BOSH director network. Ensure that service instances can access the Director so that the BOSH agents can report in.
Apps cannot access the service network. Configure Cloud Foundry application security groups to allow runtime access to the service network.
Problems accessing BOSH’s UAA or the BOSH director. Follow network troubleshooting and check that the BOSH director is online

Validate Service Broker Connectivity to Service Instances

To validate connectivity, do the following:

  1. View the BOSH deployment name for your service broker by running: 

    bosh deployments

  2. SSH into the MySQL for PCF service broker by running: 

    bosh -d DEPLOYMENT-NAME ssh

  3. If no BOSH task-id appears in the error message, look in the broker log using the broker-request-id from the task.

Validate App Access to Service Instance

Use cf ssh to access to the app container, then try connecting to the MySQL for PCF service instance using the binding included in the VCAP_SERVICES environment variable.

Quotas

Plan Quota Issues

If developers report errors such as: 

Message: Service broker error: The quota for this service plan has been exceeded.
Please contact your Operator for help.
  1. Check your current plan quota.
  2. Increase the plan quota.
  3. Log in to Ops Manager.
  4. Reconfigure the quota on the plan page.
  5. Deploy the tile.
  6. Find who is using the plan quota and take the appropriate action.

Global Quota Issues

If developers report errors such as: 

Message: Service broker error: The quota for this service has been exceeded.
Please contact your Operator for help.
  1. Check your current global quota.
  2. Increase the global quota.
  3. Log in to Ops Manager.
  4. Reconfigure the quota on the on-demand settings page.
  5. Deploy the tile.
  6. Find out who is using the quota and take the appropriate action.

Failing Jobs and Unhealthy Instances

To determine whether there is an issue with the MySQL for PCF deployment:

  1. Inspect the VMs by running: 

    bosh -d service-instance_GUID vms --vitals

  2. For additional information, run: 

    bosh -d service-instance_GUID instances --ps --vitals

If the VM is failing, follow the service-specific information. Any unadvised corrective actions (such as running BOSH restart on a VM) can cause issues in the service instance.

A failing process or failing VM might come back automatically after a temporary service outage. See VM Process Failure and VM Failure.

AZ or Region Failure

Failures at the IaaS level, such as Availability Zone (AZ) or region failures, can interrupt service and require manual restoration. See AZ Failure and Region Failure.

Techniques for Troubleshooting

Instructions on interacting with the on-demand service broker and on-demand service instance BOSH deployments, and on performing general maintenance and housekeeping tasks

Parse a Cloud Foundry (CF) Error Message

Failed operations (create, update, bind, unbind, delete) result in an error message. You can retrieve the error message later by running the cf CLI command cf service INSTANCE-NAME. 

$ cf service myservice

Service instance: myservice
Service: super-db
Bound apps:
Tags:
Plan: dedicated-vm
Description: Dedicated Instance
Documentation url:
Dashboard:

Last Operation
Status: create failed
Message: Instance provisioning failed: There was a problem completing your request.
     Please contact your operations team providing the following information:
     service: redis-acceptance,
     service-instance-guid: ae9e232c-0bd5-4684-af27-1b08b0c70089,
     broker-request-id: 63da3a35-24aa-4183-aec6-db8294506bac,
     task-id: 442,
     operation: create
Started: 2017-03-13T10:16:55Z
Updated: 2017-03-13T10:17:58Z

Use the information in the Message field to debug further. Provide this information to Support when filing a ticket.

The task-id field maps to the BOSH task ID. For more information on a failed BOSH task, use the bosh task TASK-ID.

The broker-request-guid maps to the portion of the On-Demand Broker log containing the failed step. Access the broker log through your syslog aggregator, or access BOSH logs for the broker by typing bosh logs broker 0. If you have more than one broker instance, repeat this process for each instance.

Access Broker and Instance Logs and VMs

Before following the procedures below, log in to the cf CLI and the BOSH CLI.

Access Broker Logs and VMs

You can access logs using Ops Manager by clicking on the Logs tab in the tile and downloading the broker logs.

To access logs using the BOSH CLI, do the following:

  1. Identify the on-demand broker (ODB) deployment by running the following command: 

    bosh deployments

  2. View VMs in the deployment by running the following command: 

    bosh -d DEPLOYMENT-NAME instances

  3. SSH onto the VM by running the following command: 

    bosh -d DEPLOYMENT-NAME ssh

  4. Download the broker logs by running the following command: 

    bosh -d DEPLOYMENT-NAME logs

The archive generated by BOSH includes the following logs:

Log Name Description
broker.stdout.log Requests to the on-demand broker and the actions the broker performs while orchestrating the request (e.g. generating a manifest and calling BOSH). Start here when troubleshooting.
bpm.log Control script logs for starting and stopping the on-demand broker.
post-start.stderr.log Errors that occur during post-start verification.
post-start.stdout.log Post-start verification.
drain.stderr.log Errors that occur while running the drain script.

Access Service Instance Logs and VMs

  1. To target an individual service instance deployment, retrieve the GUID of your service instance with the following cf CLI command: 

    cf service MY-SERVICE --guid

  2. To view VMs in the deployment, run the following command: 

    bosh -d service-instance_GUID instances

  3. To SSH into a VM, run the following command: 

    bosh -d service-instance_GUID ssh

  4. To download the instance logs, run the following command: 

    bosh -d service-instance_GUID logs

Run Service Broker Errands to Manage Brokers and Instances

From the BOSH CLI, you can run service broker errands that manage the service brokers and perform mass operations on the service instances that the brokers created. These service broker errands include:

To run an errand, run the following command: 

bosh -d DEPLOYMENT-NAME run-errand ERRAND-NAME

For example: 

bosh -d my-deployment run-errand deregister-broker

Register Broker

The register-broker errand does the following:

  • Registers the service broker with Cloud Controller.
  • Enables service access for any plans that are enabled on the tile.
  • Disables service access for any plans that are disabled on the tile.
  • Does nothing for any plans that are set to manual on the tile.

You should run this errand whenever the broker is re-deployed with new catalog metadata to update the Marketplace.

Plans with disabled service access are only visible to admin Cloud Foundry users. Non-admin Cloud Foundry users, including Org Managers and Space Managers, cannot see these plans.

Deregister Broker

This errand deregisters a broker from Cloud Foundry.

The errand does the following:

  • Deletes the service broker from Cloud Controller
  • Fails if there are any service instances, with or without bindings

Use the Delete All Service Instances errand to delete any existing service instances.

To run the errand, run the following command: 

bosh -d DEPLOYMENT-NAME run-errand deregister-broker

Upgrade All Service Instances

The upgrade-all-service-instances errand does the following:

  • Collects all of the service instances that the on-demand broker has registered.
  • Issues an upgrade command and deploys the a new manifest to the on-demand broker for each service instance.
  • Adds to a retry list any instances that have ongoing BOSH tasks at the time of upgrade.
  • Retries any instances in the retry list until all instances are upgraded.

When you make changes to the plan configuration, the errand upgrades all the MySQL for PCF service instances to the latest version of the plan.

If any instance fails to upgrade, the errand fails immediately. This prevents systemic problems from spreading to the rest of your service instances.

Delete All Service Instances

This errand uses the Cloud Controller API to delete all instances of your broker’s service offering in every Cloud Foundry org and space. It only deletes instances the Cloud Controller knows about. It does not delete orphan BOSH deployments.

Note: Orphan BOSH deployments do not correspond to a known service instance. While rare, orphan deployments can occur. Use the orphan-deployments errand to identify them.

The delete-all-service-instances errand does the following:

  1. Unbinds all apps from the service instances.
  2. Deletes all service instances sequentially. Each service instance deletion includes:
    1. Running any pre-delete errands
    2. Deleting the BOSH deployment of the service instance
    3. Removing any ODB-managed secrets from BOSH CredHub
    4. Checking for instance deletion failure, which results in the errand failing immediately
  3. Determines whether any instances have been created while the errand was running. If new instances are detected, the errand returns an error. In this case, VMware recommends running the errand again.

Warning: Use extreme caution when running this errand. You should only use it when you want to totally destroy all of the on-demand service instances in an environment.

To run the errand, run the following command: 

bosh -d service-instance_GUID delete-deployment

Detect Orphaned Service Instances

A service instance is defined as “orphaned” when the BOSH deployment for the instance is still running, but the service is no longer registered in Cloud Foundry.

The orphan-deployments errand collates a list of service deployments that have no matching service instances in Cloud Foundry and return the list to the operator. It is then up to the operator to remove the orphaned BOSH deployments.

To run the errand, run the following command: 

bosh -d DEPLOYMENT-NAME run-errand orphan-deployments

If orphan deployments exist—The errand script does the following:

  • Exit with exit code 10
  • Output a list of deployment names under a [stdout] header
  • Provide a detailed error message under a [stderr] header

For example: 

[stdout]
[{"deployment\_name":"service-instance\_80e3c5a7-80be-49f0-8512-44840f3c4d1b"}]

[stderr]
Orphan BOSH deployments detected with no corresponding service instance in Cloud Foundry. Before deleting any deployment it is recommended to verify the service instance no longer exists in Cloud Foundry and any data is safe to delete.

Errand 'orphan-deployments' completed with error (exit code 10)

These details will also be available through the BOSH /tasks/ API endpoint for use in scripting: 

$ curl 'https://bosh-user:bosh-password@bosh-url:25555/tasks/task-id/output?type=result' | jq .
{
  "exit_code": 10,
  "stdout": "[{"deployment_name":"service-instance_80e3c5a7-80be-49f0-8512-44840f3c4d1b"}]\n",
  "stderr": "Orphan BOSH deployments detected with no corresponding service instance in Cloud Foundry. Before deleting any deployment it is recommended to verify the service instance no longer exists in Cloud Foundry and any data is safe to delete.\n",
  "logs": {
    "blobstore_id": "d830c4bf-8086-4bc2-8c1d-54d3a3c6d88d"
  }
}

If no orphan deployments exist—The errand script does the following:

  • Exit with exit code 0
  • Stdout will be an empty list of deployments
  • Stderr will be None
[stdout]
[]

[stderr]
None

Errand 'orphan-deployments' completed successfully (exit code 0)

If the errand encounters an error during running—The errand script does the following:

  • Exit with exit 1
  • Stdout will be empty
  • Any error messages will be under stderr

To clean up orphaned instances, run the following command on each instance:

WARNING: Running this command may leave IaaS resources in an unusable state. 

bosh delete-deployment service-instance_SERVICE-INSTANCE-GUID

Retrieve Admin and Read-Only Admin Credentials for a Service Instance

To retrieve the admin credentials for a service instance from BOSH CredHub:

  1. Use the cf CLI to determine the GUID associated with the service instance for which you want to retrieve credentials by running: 
    cf service SERVICE-INSTANCE-NAME --guid
    For example: 
    $ cf service my-service-instance --guid

12345678-90ab-cdef-1234-567890abcdef
    If you do not know the name of the service instance, you can list service instances in the space with cf services.
  2. Follow the steps in Gather Credential and IP Address Information and Log In to the Ops Manager VM with SSH of Advanced Troubleshooting with the BOSH CLI to SSH into the Ops Manager VM.
  3. From the Ops Manager VM, log in to your BOSH Director with the BOSH CLI. See Authenticate with the BOSH Director VM in Advanced Troubleshooting with the BOSH CLI.

  4. Find the values for BOSH_CLIENT and BOSH_CLIENT_SECRET:

    1. In the Ops Manager Installation Dashboard, click the BOSH Director tile.
    2. Click the Credentials tab.
    3. In the BOSH Director section, click the link to the BOSH Commandline Credentials .
    4. Record the values for BOSH_CLIENT and BOSH_CLIENT_SECRET.
  5. Set the API target of the CredHub CLI to your BOSH CredHub server by running: 
    credhub api https://BOSH-DIRECTOR-IP:8844 \
      --ca-cert=/var/tempest/workspaces/default/root_ca_certificate
    Where BOSH-DIRECTOR-IP is the IP address of the BOSH Director VM.

    For example: 
    $ credhub api https://10.0.0.5:8844 \
      --ca-cert=/var/tempest/workspaces/default/root_ca_certificate
  6. Log in to CredHub by running: 
    credhub login \
    --client-name=BOSH-CLIENT \
    --client-secret=BOSH-CLIENT-SECRET

    For example:

    $ credhub login \
      --client-name=credhub \
      --client-secret=abcdefghijklm123456789

  7. Use the CredHub CLI to retrieve the credentials by doing one of following :

    • Retrieve the password for the admin user by running: 
      credhub get -n /p-bosh/service-instance_GUID/admin_password
      In the output, the password appears under value. Record the password.
      For example: 
      $ credhub get \
  -n /p-bosh/service-instance_70d30bb6-7f30-441a-a87c-05a5e4afff26/admin_password 
      
  id: d6e5bd10-3b60-4a1a-9e01-c76da688b847
  name: /p-bosh/service-instance_70d30bb6-7f30-441a-a87c-05a5e4afff26/admin_password
  type: password
  value: UMF2DXsqNPPlCNWMdVMcNv7RC3Wi10
  version_created_at: 2018-04-02T23:16:09Z
    • Retrieve the password for the read-only admin user by running: 
      credhub get -n /p-bosh/service-instance_GUID/read_only_admin_password
      In the output, the password appears under value. Record the password.
  8. Record the IP of your service instance. See Connect Using an IP Address.
  9. Connect to your database by doing one of following:
    • Connect using a management tool. See Using Management Tools for MySQL for PCF.
    • Connect directly from your workstation using the MySQL client by running: 
      mysql -h IP-ADDRESS -u admin -P 3306 -p
      When prompted for a password, enter the password you recorded.

Reinstall a Tile

To reinstall the MySQL for PCF tile, see Reinstalling MySQL for Pivotal Cloud Foundry version 2 and above in the Pivotal Support knowledge base.

View Resource Saturation and Scaling

To view usage statistics for any service, do the following:

  1. Run the following command: 

    bosh -d DEPLOYMENT-NAME vms --vitals

  2. To view process-level information, run: 

    bosh -d DEPLOYMENT-NAME instances --ps

Identify Apps using a Service Instance

To identify which apps are using a specific service instance from the name of the BOSH deployment:

  1. Take the deployment name and strip the service-instance_ leaving you with the GUID.
  2. Log in to CF as an admin.
  3. Obtain a list of all service bindings by running the following: 
    cf curl /v2/service_instances/GUID/service_bindings
  4. The output from the above curl gives you a list of resources, with each item referencing a service binding, which contains the APP-URL. To find the name, org, and space for the app, run the following:
    1. cf curl APP-URL and record the app name under entity.name.
    2. cf curl SPACE-URL to obtain the space, using the entity.space_url from the above curl. Record the space name under entity.name.
    3. cf curl ORGANIZATION-URL to obtain the org, using the entity.organization_url from the above curl. Record the organization name under entity.name.

Note: When running cf curl ensure that you query all pages, because the responses are limited to a certain number of bindings per page. The default is 50. To find the next page curl the value under next_url.

Monitor Quota Saturation and Service Instance Count

Quota saturation and total number of service instances are available through ODB metrics emitted to Loggregator. The metric names are shown below:

Metric Name Description
on-demand-broker/SERVICE-NAME-MARKETPLACE/quota_remaining global quota remaining for all instances across all plans
on-demand-broker/SERVICE-NAME-MARKETPLACE/PLAN-NAME/quota_remaining quota remaining for a particular plan
on-demand-broker/SERVICE-NAME-MARKETPLACE/total_instances total instances created across all plans
on-demand-broker/SERVICE-NAME-MARKETPLACE/PLAN-NAME/total_instances total instances created for a given plan

Note: Quota metrics are not emitted if no quota has been set.

Techniques for Troubleshooting Highly Available Clusters

If your cluster is experiencing downtime or in a degraded state, Pivotal recommends gathering information to diagnose the type of failure the cluster is experiencing with the following workflow:

  1. Consult solutions for common errors. See Highly Available Cluster Troubleshooting Errors above.
  2. Use mysql-diag to view a summary of the network, disk, and replication state of each cluster node. Depending on the output from mysql-diag, you might recover your cluster with the following troubleshooting techniques: For more information about mysql-diag, see Running mysql-diag.
  3. Run download-logs against each node in your MySQL for PCF cluster, proxies, and jumpbox VM. You must run download-logs before attempting recovery because any failures in the recovery procedure can result in logs being lost or made inaccessible.
    For more information, see the download-logs section below.

    Note: Pivotal recommends that you use the -X flag to get the complete set of available logs. However, if your cluster processes a high volume of transactions, the complete set might be too large and you can omit this flag to fetch the essential set of logs.

  4. If you are uncertain about the recovery steps to take, submit a ticket through Pivotal Support. When you submit a ticket provide the following information:
    • mysql-diag output: A summary of the network, disk, and replication state. The Running mysql-diag topic explains how to run mysql-diag.
    • download-logs logs: Logs from your MySQL for PCF cluster, proxies, and jumpbox VM. The download-logs section below explains how to run download-logs.
    • Deployment environment: The environment that MySQL for PCF is running in such as Pivotal Application Service or a service tile.
    • Version numbers: The versions of the installed Ops Manager, PAS, and MySQL for PCF.

Warning: Do not attempt to resolve cluster issues by reconfiguring the cluster, such as changing the number of nodes or networks. Only follow the diagnosis steps in this document. If you are unsure how to proceed, contact Pivotal Support.

Force a Node to Rejoin a Highly Available Cluster Manually

If a detached node fails to rejoin the cluster after a configured grace period, you can manually force the node to rejoin the cluster. This procedure removes all the data on the node, forces the node to join the cluster, and creates a new copy of the cluster data on the node.

Warning: If you manually force a node to rejoin the cluster, data stored on the local node is lost. Do not force nodes to rejoin the cluster if you want to preserve unsynchronized data. Only do this procedure with the assistance of Pivotal Support.

Before following this procedure, try to bootstrap the cluster. For more information, see Bootstrapping.

To manually force a node to rejoin the cluster, do the following:

  1. SSH into the node by following the procedure in BOSH SSH.
  2. Become root by running: 
     sudo su
  3. Shut down the mysqld process on the node by running: 
      monit stop galera-init
  4. Remove the unsynchronized data on the node by running: 
      rm -rf /var/vcap/store/pxc-mysql
  5. Prepare the node before restarting by running: 
      /var/vcap/jobs/pxc-mysql/bin/pre-start
  6. Restart the mysqld process by running: 
     monit start galera-init

Re-create a Corrupted VM in a Highly Available Cluster

To re-create a corrupted VM:

  1. To log in to the BOSH Director VM by doing the following procedures:
    1. Gather the information needed to log in to the BOSH Director VM by doing the procedure in Gather Credential and IP Address Information.
    2. Log in to the Ops Manager VM by doing the procedure in Log in to the Ops Manager VM with SSH.
    3. Log in to the BOSH Director VM by doing the procedure in Log in to the BOSH Director VM.
  2. Identify and re-create the unresponsive node with bosh cloudcheck, by doing the procedure in BOSH Cloudcheck and run Recreate VM using last known apply spec.

    Warning: Recreating a node will clear its logs. Ensure the node is completely down before recreating it.

    Warning: Only re-create one node. Do not re-create the entire cluster. If more than one node is down, contact Pivotal Support.

Check Replication Status in a Highly Available Cluster

If you see stale data in your cluster, you can check whether replication is functioning normally.

To check the replication status, do the following:

  1. To log in to the BOSH Director VM, do the following:
    1. Gather the information needed to log in to the BOSH Director VM by doing the procedure in Gather Credential and IP Address Information.
    2. Log in to the Ops Manager VM by doing the procedure in Log in to the Ops Manager VM with SSH.
  2. Create a dummy database in the first node by running: 
    mysql -h FIRST-NODE-IP-ADDRESS \
       -u YOUR-IDENTITY \
       -p -e "create database verify_healthy;"
    Where:
    • FIRST-NODE-IP-ADDRESS is the IP address of the first node you recorded in step 1.
    • YOUR-IDENTITY is the value of identity that you recorded in step 1.
  3. Create a dummy table in the dummy database by running: 
    mysql -h FIRST-NODE-IP-ADDRESS \
      -u your-identity \
      -p -D verify_healthy \
      -e "create table dummy_table (id int not null primary key auto_increment, info text) \
      engine='innodb';"
  4. Insert data into the dummy table by running: 
    mysql -h FIRST-NODE-IP-ADDRESS \
      -u YOUR-IDENTITY \
      -p -D verify_healthy \
      -e "insert into dummy_table(info) values ('dummy data'),('more dummy data'),('even more dummy data');"

  5. Query the table and verify that the three rows of dummy data exist on the first node by running: 

    mysql -h FIRST-NODE-IP-ADDRESS \
      -u YOUR-IDENTITY \
      -p -D verify_healthy \
      -e "select * from dummy_table;"
    When prompted for a password, provide the password value recorded in step 1.
    The above command returns output similar to the following: 
      +----+----------------------+
  | id | info                 |
  +----+----------------------+
  |  4 | dummy data           |
  |  7 | more dummy data      |
  | 10 | even more dummy data |
  +----+----------------------+
  6. Verify that the other nodes contain the same dummy data by doing the following for each of the remaining MySQL server IP addresses:
    1. Query the dummy table by running : 
      mysql -h NEXT-NODE-IP-ADDRESS \
          -u YOUR-IDENTITY \
          -p -D verify\_healthy \
          -e "select * from dummy_table;"
      When prompted for a password, provide the password value recorded in step 1.
    2. Verify that the node contains the same three rows of dummy data as the other nodes by running: 
      mysql -h NEXT-NODE-IP-ADDRESS \
          -u YOUR-IDENTITY \
          -p -D verify\_healthy \
          -e "select * from dummy\_table;"
      When prompted for a password, provide the password value recorded in step
    3. Verify that the above command returns output similar to the following: 
          +----+----------------------+
    | id | info                 |
    +----+----------------------+
    |  4 | dummy data           |
    |  7 | more dummy data      |
    | 10 | even more dummy data |
    +----+----------------------+
  7. If each MySQL server instance does not return the same result, before proceeding further or making any changes to your deployment, contact Pivotal Support
    If each MySQL server instance returns the same result, then you can safely proceed to scaling down your cluster to a single node.

Tools for Troubleshooting

The troubleshooting techniques described above use the following tools.

download-logs

download-logs is a script that you can run from your Ops Manager VM to aggregate logs from your MySQL for PCF cluster nodes, proxies, and, with highly available clusters, the jumpbox VM.

To use the download-logs script:

  1. Download and unzip the download-logs script from MySQL for PCF on Pivotal Network.

  2. From the Ops Manager Installation Dashboard, navigate to BOSH Director > Credentials.

  3. Click Link to Credential for the Bosh Commandline Credentials.

  4. From the plaintext file that opens, record the values for the following:

    • BOSH_CLIENT
    • BOSH_CLIENT_SECRET
    • BOSH_CA_CERT
    • BOSH_ENVIRONMENT

  5. From the BOSH CLI, view the name of the BOSH deployment for MySQL for PCF by running:

    bosh deployments

    Record the name of the BOSH deployment.

  6. SSH into your Ops Manager VM by doing the procedures in Gather Credential and IP Address Information and SSH into Ops Manager.

  7. File transfer or copy-paste the download-logs script to a working directory on the Ops Manager VM.

  8. Set local environment variables to the same BOSH variable values that you recorded earlier, including BOSH_DEPLOYMENT for the deployment name.

    For example: 

    $ BOSH_CLIENT=ops_manager \
  BOSH_CLIENT_SECRET=a123bc-E_4Ke3fb-gImbl3xw4a7meW0rY
  BOSH_CA_CERT=/var/tempest/workspaces/default/root_ca_certificate \
  BOSH_ENVIRONMENT=10.0.0.5 \
  BOSH_DEPLOYMENT=pivotal-mysql-14c4

  9. Run the download-logs script by running:

    ./download-logs -o .

    The script saves a compressed file of logs combined from all MySQL for PCF VMs. The filename has the form TIMESTAMP-mysql-logs.tar.gz.gpg.

mysql-diag

The mysql-diag tools outputs the current status of a highly available (HA) MySQL for PCF cluster in PCF and suggests recovery actions if the cluster fails.

For more information, see Running mysql-diag.

Knowledge Base (Community)

Find the answer to your question and browse product discussions and solutions by searching the VMware Tanzu Knowledge Base.

File a Support Ticket

You can file a ticket with Support. Be sure to provide the error message from cf service YOUR-SERVICE-INSTANCE.

To expedite troubleshooting, provide your service broker logs and your service instance logs. If your cf service YOUR-SERVICE-INSTANCE output includes a task-id, provide the BOSH task output.