It takes too long to init service databases

[aeva black]

The various calls to "init-db" take far, far longer than I think they should. See
  https://review.openstack.org/#/c/50745/
for an example of the lengths being taken to accommodate this. Something must be done besides sleep($more)!

My suspicion is that a few well-placed config changes will drastically reduce the time needed to run all these ALTERs on a (mostly) empty data set. This could be adjusting innodb's file-per-table or flush-log-at-trx-commit settings, or mysql's open_files_limit or table_open_cache or sync_binlog settings, or some tuning of the FS to help with performance of many rapid metadata changes (eg, mount with noatime and nodiratime, or change xfs journal size if using xfs). Or something else.

[Gregory Haynes (greghaynes)]

Im fairly certain this has been "fixed" by default now. As evidenced by http://logs.openstack.org/66/148766/2/check-tripleo/check-tripleo-ironic-undercloud-precise-nonha/62791c2/logs/seed_logs/os-collect-config.txt.gz#_Jan_21_01_11_38 where migrations take a couple seconds (and most of that time is just getting to the point where migrations start running).