Cinder - Metrics & Alerts (LVM and Ceph)

asked 2016-05-06 06:12:03 -0500

mickm gravatar image

I'm writing some Nagios checks / Wavefront Metrics for Cinder. The idea being to monitor the storage cluster and raise alerts for certain metrics falling below expected thresholds - so far I have the following ideas for checks:

i) Cinder backed by LVM

  • A Nagios check for volume information - e.g. number of Cinder volumes. The check runs a MySQL check on Cinder database and returns the amount of volumes in i) available ii) deleted or iii) in-use status, i.e.:

    mysql> SELECT status, count(*) FROM volumes GROUP BY status;

    +-----------+----------+

    | status | count(*) |

    +-----------+----------+

    | available | 9 |

    | deleted | 1482 |

    | in-use | 15 |

    +-----------+----------+

    3 rows in set (0.01 sec)

  • Cinder quota information: A check running the cinder-volume-usage-audit command will display the total number of volumes in use: e.g. output: Found 691 volumes. While this is relevant info in itself. Something more relevant might be the overall quota numbers - e.g. there are x volumes being used out of a quota of y. An alert then being raised when a certain threshold has been reached/exceeded. Unfortunately Cinder CLI only supports per tenant quotas vs. overall cluster quotas, e.g. usage:

cinder quota-show <tenant-id> cinder quota-usage <tenant-id>

Is anyone aware of a way to achieve an overall quota metric for a cluster?

  • lvdisplay: perhaps lvdisplay could be used to show how much data is available / being used for LVM volumes?

ii) Cinder backed by Ceph

I'm notyet sure which metrics would be useful here.

Please suggest any ways you might improve my current metrics/alerts.

Thanks.

edit retag flag offensive close merge delete