{
"build_id": "2014-11-17_17-53-34",
"ostf_sha": "82465a94eed4eff1fc8d8e1f2fb7e9993c22f068",
"build_number": "504",
"auth_required": true,
"api": "1.0",
"nailgun_sha": "8d23d1b1bcd9213a70a40c38c3c1486d215d40b5",
"production": "docker",
"fuelmain_sha": "8d4943d5ead7a894d4af5e10172510fa60eeed84",
"astute_sha": "65eb911c38afc0e23d187772f9a05f703c685896",
"feature_groups": [
"mirantis"
],
"release": "6.0",
"release_versions": {
"2014.2-6.0": {
"VERSION": {
"build_id": "2014-11-17_17-53-34",
"ostf_sha": "82465a94eed4eff1fc8d8e1f2fb7e9993c22f068",
"build_number": "504",
"api": "1.0",
"nailgun_sha": "8d23d1b1bcd9213a70a40c38c3c1486d215d40b5",
"production": "docker",
"fuelmain_sha": "8d4943d5ead7a894d4af5e10172510fa60eeed84",
"astute_sha": "65eb911c38afc0e23d187772f9a05f703c685896",
"feature_groups": [
"mirantis"
],
"release": "6.0",
"fuellib_sha": "8a0ceff90777af75a3f9363a57185e608f3ee10d"
}
}
},
"fuellib_sha": "8a0ceff90777af75a3f9363a57185e608f3ee10d"
}
Steps:
1. Create and deploy next cluster - Ubuntu, HA, Neutron GRE, Image-based provisioning, 3 controller, 2 compute, 1 cinder node
2. After deployment delete cluster
3. Create new cluster - CentOS, HA, Neutron Vlan, Image-based provisioning, 3 controller, 2 compute nodes
4. Provision cluster
Expected - all nodes were successfully provisioned
Actual - 1 time out of 4 one of the nodes is provisioned but after node's restart and start up of cloud init it destroyed all node's files
Logs are attached
The root cause is still unknown. Stable and repeatable way of how to reproduce it even.
At first i thought it was a failure inside of the boothook script. Nope, the boothook script worked fine every time i'd tried. https:/
It might be cloud-init's semaphores issues. For unknown reason cloud-init log was full of messages showing that all config_* modules have been already run. The executor simply checks the semaphore and skips if it exists.
They stored in /var/lib/
I have only one strategy to follow. We need to disable automatic cloud-init start on boot (just removing links from /etc/rc.d/* should help) and then start cloud-init by hand under `strace` or other hardcore debug stuff and watch what will happen.
Sounds as very time consuming task.