Time desync on slaves leads to Cinder errors
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
Fix Released
|
Critical
|
Stanislaw Bogatkin | ||
6.0.x |
Invalid
|
Undecided
|
Stanislaw Bogatkin |
Bug Description
api: '1.0'
astute_sha: ed5270bf9c6c123
auth_required: true
build_id: 2015-01-29_10-46-25
build_number: '85'
feature_groups:
- mirantis
fuellib_sha: b2fbaa9ffb74faf
fuelmain_sha: ''
nailgun_sha: 648cad955ec88af
ostf_sha: c9100263140008a
production: docker
python-
release: '6.1'
Smoke tests failed (Ubuntu, NeutronGRE, CinderLVM) because OSTF was unable to create volume:
Environment details:
- Ubuntu, Simple (multinode), NeutronVlan
- 1 controller and 2 compute+cinder nodes
I reproduced the same issue with creating volume manually:
http://
Cinder was unable to create new volume because all cinder-volume services were down:
http://
I found that time on slaves is desynced:
http://
And NTP daemon isn't running:
http://
After time synchronization on all slaves (I ran "ntpdate -vu $(awk '/^server/ && $2 !~ /127.*/ {print $2}' /etc/ntp.conf)" command), issue with cinder volumes was fixed.
Changed in fuel: | |
assignee: | Fuel Library Team (fuel-library) → Stanislaw Bogatkin (sbogatkin) |
status: | New → Confirmed |
Changed in fuel: | |
status: | Fix Committed → Fix Released |
Ubuntu HA should be vulnerable to that bug too. etc/ntp. conf && command'. As long as package ntp-dev installed before snippet applies, so /var/lib/ntp alreade exists on moment applying that part of snippet and command that adding string with 'server' doesn't applies. After that ubuntu node sync to nothing and after that time is desynced.
It's happened because new ntp 4.2.8 package was merged into master, and it create /var/lib/ntp directory. In our ntp preseed snippet we use long line of 'command && mkdir /var/lib/ntp && command && echo "server $server burst iburst" >> /target/
Workaround is manually created string "server <server name or IP here> burst iburst" in /etc/ntp.conf.