When one wants to migrate data from one node to another in order to decomission a node, theres several steps involved. However it involves temporarily offlining data. 1. replace-brick to migrate data from one node to another. 2. remove-brick to remove original brick location. 3. add-brick to add the new brick location. In step 2, theres downtime. According to what we advertise, customer would be led to believe this is online migration. If we pointed the bricks in the volume automatically after the replace-brick command, it would make everything easier and hopefully without the need to issue two more command and allow data to remain online. Question is are these the right steps or was replace-brick meant to point to the new location?
If one wants to migrate data from one node to another in order to decommission a node, the steps involved are - 1. replace-brick old:/path new:/path 2. replace-brick status (to monitor progress) 3. replace-brick commit (after completion, or replace-brick abort) There is no temporary offline of data.
This is more of a lack of documentation than invalid. How does this get to our web site or documentation?
No commit option volume replace-brick <VOLNAME> (<BRICK> <NEW-BRICK>) start|pause|abort|status - replace-brick operations
Documentation needs to be fixed to show the 'commit' option
I have added information on commit option at: http://www.gluster.com/community/documentation/index.php/Gluster_3.1:_Migrating_Volumes Regards, Divya
Its fixed now. The link provided by Divya included information about migation start, pause, status, abort and commit.