This can only be caused by running Galera in multiwriter mode (probably unintentionally, due to https://bugs.launchpad.net/fuel/+bug/1490595).
Anyway, this should be a really rare case, as usually instance_info_cache_update() is called only from 1 nova-compute at a time, except for the case when it can be updated from nova-api on association / disassociation of a floating IP.
This can only be caused by running Galera in multiwriter mode (probably unintentionally, due to https:/ /bugs.launchpad .net/fuel/ +bug/1490595).
Anyway, this should be a really rare case, as usually instance_ info_cache_ update( ) is called only from 1 nova-compute at a time, except for the case when it can be updated from nova-api on association / disassociation of a floating IP.