tripleo-transfer ansible module cannot customize some transfer flags
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
tripleo |
Fix Released
|
Medium
|
Damien Ciabrini |
Bug Description
We use tripleo-transfer during FFU to copy the contents of the most up-to-date DB to the controller node that has just been upgraded, so it catches up before being restarted.
Currently the transfer module applies some default rsync transfer options [1], which are not optimal for dealing with the database transfer:
- delta-transfer is used, which requires hashing the contents of the entire DB, which is fairly time consuming.
- transfer compression is applied, which is often time not effective with the DB contents, thus more time consuming
- no in-place copy is attempted, which increases the storage requirement on the destination.
- all the files from the sources are copied to the destination, which is not idea because we can't skip useless or potentially unwanted files.
For performance and reliability, all the points above could be enhanced so that the transfer would look more like what galera does when transferring the full DB during an SST: in-place, unconditional copy (no checksum), no compression, selective copy.
[1] --delay-updates -F --ignore-times --compress --archive --delete
Changed in tripleo: | |
milestone: | wallaby-rc1 → xena-1 |
Changed in tripleo: | |
milestone: | xena-1 → xena-2 |
Fix proposed to branch: master /review. opendev. org/c/openstack /tripleo- heat-templates/ +/787326
Review: https:/