I believe it was the Ironic-conductor service from OpenStack. I haven't confirmed it yet, but I stopped all Ironic containers after a reboot and I am just waiting now to see if the machine slows down. This problem migrated to the other nodes as well and when I stopped the ironic containers on one of them the used RAM went way down.
I haven't confirmed this though, but I did learn how to troubleshoot this issue so I marked the problem as solved.
|