[rhos-list] migrations with gluster isue

[Paul Robert Marino]

hello every one

Ive been doing some testing with RDO grizzly and found an issue with
the compute nodes doing a migration.
Live migration works as expected, but regular migrations don't. I
wouldn't be too worried about it except for the migrate instance
option in the admin tab in horizon.
for some reason they don't seem to detect that they are on shared
storage and instead try to ssh to each other and do filesystem
modifications.
the first one it tries is an mkdir then it fails right now ive left
the shell for the nova user as /sbin/nologin so even if i did add keys
it wouldnt work and besides the mkdir command would return an error
any way.


here are the settings ive been tinkering with
"
instances_path = /nova-cluster/instances
live_migration_bandwidth=0
live_migration_flag=VIR_MIGRATE_UNDEFINE_SOURCE,VIR_MIGRATE_PEER2PEER,VIR_MIGRATE_LIVE
live_migration_retry_count=30
live_migration_uri=qemu+tcp://%s/system
image_cache_manager_interval=0
block_migration_flag=VIR_MIGRATE_UNDEFINE_SOURCE,VIR_MIGRATE_PEER2PEER
glusterfs_mount_point_base=/nova-cluster
"

does any one have any ideas about this or know of any settings I may
have missed?