The core of the issue is that Fuel always runs the post_deployment task even though the operator asked for the re-execution of only one task.
Regarding lma_collector, the problem is in the heka Puppet module [1]. It tries to stop heka running /etc/init.d/heka which will kill any process named hekad.
Regarding collectd, it is more tricky: the lma_collector/base.pp manifest wipes out the conf.d directory to get rid of the default configuration and the other manifests (controller.pp, compute.pp, ...) add on top of that.
The core of the issue is that Fuel always runs the post_deployment task even though the operator asked for the re-execution of only one task.
Regarding lma_collector, the problem is in the heka Puppet module [1]. It tries to stop heka running /etc/init.d/heka which will kill any process named hekad.
Regarding collectd, it is more tricky: the lma_collector/ base.pp manifest wipes out the conf.d directory to get rid of the default configuration and the other manifests (controller.pp, compute.pp, ...) add on top of that.
[1] https:/ /github. com/openstack/ fuel-plugin- lma-collector/ blob/96df47af73 85578280a68863b 55416da7dd8bfdb /deployment_ scripts/ puppet/ modules/ heka/manifests/ init.pp# L111