Revision 4a402f74
| ID | 4a402f74e1670dc642e2345901ded87b8bb144db |
| Parent | 228d0df5 |
| Child | 7062e77d |
rework premigrate and postmigrate scripts
The change was inspired by the live-migrate scripts for SYSTEM_DS with `shared` TM_MAD
After some rework I've figured out that it could be made as some sort of generic one.
The proposed changes- tm_common.sh - new function migrate_other() receiving all arguments from the caller script
The function parses the TEMPLATE blob and exposes BASH arrays:
DISK_ID_ARRAY - array of all VM DISK_ID
TM_MAD_ARRAY - array of TM_MADs for each DISK_ID
CLONE_ARRAY - aray of all CLONE values for each DISK_ID
The function exposes DISK_ID of the CONTEXT disk too as CONTEXT_DISK_ID
The function check for extra argument to prevent recursion loops
For each TM_MAD it it is not current one the corresponding script is called
with same arguments but one additional - to mark that it is called not from
in the SYSTEM_DS context
Usage:
migrate_common "$@"The function is appended to the following TM_MADs:
ceph
fs_lvm
qcow2
shared
ssh
The General idea is to allow live-migration of VMs with mix of disks with any of the above TM_MADs (different datastores)
For example if we have VM with the following disks
disk.0 ceph
disk.1 ceph (context)
disk.2 - ceph
disk.3 - fs_lvm
disk.4 - qcow2
In the above scenario when live migration is issed the following scripts will be called
ceph/premigrate <args>
fs_lvm/premigrate <args> "ceph"
qcow2/premigrate <args> "ceph"
As most probably I am missing someting I am open for discussion :)
Files
- added
- modified
- copied
- renamed
- deleted