I asked Maxim for ssh credentials and logged in to the env. I added some debug prints around some code and it seems that everywhere where memcache does something, it may lead to lags. For example, http://paste.openstack.org/show/193705/ from keystone/common/kvs/backends/memcached.py leads to http://paste.openstack.org/show/193704/ (please note the times).
Sometimes there is a weird gap between releasing connections in memcache pool and start of acquiring of some lock: http://paste.openstack.org/show/193706/ (note the times).
Anyway, there is no much sense in steps 3 and 4 now.
I asked Maxim for ssh credentials and logged in to the env. I added some debug prints around some code and it seems that everywhere where memcache does something, it may lead to lags. For example, http:// paste.openstack .org/show/ 193705/ from keystone/ common/ kvs/backends/ memcached. py leads to http:// paste.openstack .org/show/ 193704/ (please note the times).
Sometimes there is a weird gap between releasing connections in memcache pool and start of acquiring of some lock: http:// paste.openstack .org/show/ 193706/ (note the times).
Anyway, there is no much sense in steps 3 and 4 now.