I have 6 nodes that have all been commissioned. I can ssh to 2 of them with ubuntu@ip_address. The other 4 fail to ssh with
Permission denied (publickey)
Juju deploy fail with
Failed to query node's BMC — Node could not be queried node-3c9e137a-744c-11e4-a75c-180373b04ac9 (esxi10.local) connection timeout.
I have added ntp
runs during startup and via cron
via commissioning scripts. I have even tried adding a second account, complete with password and public keys, but I can't log into these either.
Why would some machines commission correctly and others not? Any ideas?
The problem appears to have been caused by the physical clocks being not in sync enough with the server and critical parts of the commissioning happen before they run
ntpdate
. I installed 3 commissioning files to syncntpdate
and the hardware clock.01-ntpdate
- syncs withntpdate
and then writes to the hardware clock01-ntpdate-cron
- installs acron
job to sync type every 15 minutes and set thehwclock
01-S10ntpdate
- installs a boot up script to sync time (and hwclock) on bootupAfter adding these 3 commissioning files, the first commissioning installs them and actually sets the
hwclock
(and usually fails the commissioning because the fails before it actually sets the time correctly). The next commissioning runs successfully.Also, set up a local ntp server, and in /etc/maas/preseeds/preseed-master look for: d-I clock-setup/ntp-server string ntp.ubuntu.com and change ntp.ubuntu.com to your ntp server's IP or machine name.