Open Stack

Hi

We're using Ceilometer with Gnocchi in Rocky release. And we're using
cpu_util variable to doing auto scaling on Heat.

We known that ceilometer will create its own archive policy in gnocchi, and
default value of granularity is 5 minutes. We want change
ceilometer's granularity to a minute. Here's our workaround.

1. Edit pipeline.yaml. Since gnocchi has default policy that granularity =
1 second, we change "gnocchi://" to "gnocchi://?archive_policy=high"

2. Edit ceilometer.conf & polling.yaml, change both evaluation_interval
(ceilometer.conf) & interval (polling.yaml) to 60.

3. Restart all ceilometer services

We tested this workaround but found that it would not work sometimes. And
found that the first detection triggered every "o'clock" when created auto
scaling at first time.

If we want to shrink the granularity time, is our workaround correct? And
how is the correct way to do this if not?


Many thanks,
Eddie.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-discuss/attachments/20191025/1a182bc4/attachment.html>