Announcement

Collapse

New forum rules

The forum get new posts rules.

A new user will be moderate for new post. After five valid posts, the user will be to have right to post without moderation.

If you have some issue to post or to connect, contact us.
See more
See less

[18.10] perf data for new machines are not visible

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • [18.10] perf data for new machines are not visible

    Hello,

    In an environment migrated from 2.8 to 18.10, we can't get new VMs' performance metrics.
    For VMs that existed before the migration, the data can be viewed. But for new ones, nothing is visible.
    In a clean install of Centreon 18.10, we don't have this issue.

    From what I understand, the performance data should be written into the "index_data", "metrics" and "data_bin" tables.
    We can see the index_data table contains the expected entries. But the "metrics" table is not filled - no data for the expected machine. The data_bin table contains something, but it doesn't seem to be updated as often as it should (now this may be an optimization - I'm not sure how often it should be written).

    Also the RRD files seem to be written properly.

    When we enable full logs, we can see the metrics taken into account, but no message shows that it's been written to the table.

    Code:
    [1552490956] debug: storage: processing service status event of service 241 of host 1 (ctime 1552490952)
    [1552490956] debug: storage: found index 8 of (1, 241) in cache
    [1552490956] debug: storage: generating status event for (1, 241) of index 8
    [1552490956] debug: storage: parsing perfdata string '/=6663MB;13467;15150;0;16834'
    [1552490956] debug: storage: got new perfdata (name=/, value=6663.000000, unit=MB, warning=13467.000000, critical=15150.000000, min=0.000000, max=16834.000000)
    [1552490956] debug: storage: found metric 16 of (8, /) in cache
    [1552490956] info: storage: updating metric 16 of (8, /) (unit: MB, warning: 0.000000:13467.000000, critical: 0.000000:15150.000000, min: 0.000000, max: 16834.000000)
    [1552490956] debug: storage: generating perfdata event for metric 16 (name /, ctime 1552490952, value 6663.000000)
    [1552490956] debug: storage: 2 have not yet been acknowledged
    [1552490956] debug: multiplexing: acknowledging 0 events from central-broker-master-perfdata event queue
    [1552490956] debug: failover: reading event from multiplexing engine for endpoint 'central-broker-master-perfdata'
    [1552490956] debug: failover: writing event of multiplexing engine to endpoint 'central-broker-master-perfdata'
    [1552490956] debug: storage: processing service status event of service 518 of host 159 (ctime 1552490954)
    [1552490956] debug: storage: found index 617 of (159, 518) in cache
    [1552490956] debug: storage: generating status event for (159, 518) of index 617
    [1552490956] debug: storage: parsing perfdata string 'procs=2;1:;1:;0;'
    [1552490956] debug: storage: got new perfdata (name=procs, value=2.000000, unit=, warning=inf, critical=inf, min=0.000000, max=nan)
    [1552490956] debug: storage: found metric 330 of (617, procs) in cache
    [1552490956] debug: storage: generating perfdata event for metric 330 (name procs, ctime 1552490954, value 2.000000)
    [1552490956] debug: storage: 3 have not yet been acknowledged

    From what I find in the centreon-broker code, the data in the "metrics" table is written there when a new metrics shows up that's not found in the cache. But the cache itself is created based on the content of the "metrics" table. Now the log shows that all the metrics taken into account are found in the cache - how can it be, given that the metrics table doesn't have anything for the selected VM ?
    I am sure that I am missing something - is there a way to force the metrics to be written in the "metrics" table ?


    Anything else we should look at to troubleshoot this issue ?

Working...
X