Comment 3 for bug 1016633

OK, so I grepped out the "instance:" lines for a 4-node run that took more than a minute and a 1-node run that took 10 seconds to create five files that can be compared. I noticed several strange things for the 4-node logs.

First, there was a 14 second hiccup on all four between "Starting instance..." and "Instance network info due to a call to heal_instance_info_cache. Since this fires every minute I assume it is not supposed to take 14 seconds.

Just 2 of the 4, and not the one, had an "iptables firewall:" line, and one of them took 17 seconds to get to the next thing.

I am attaching the files.