[Gluster-users] about tail command

[Alan Millar]

>>  1. the command "gluster v replace-brick " is async or sync?  The 
> replace is
>>  complete when the command quit ?
> It is a sync command, replacing the brick finishes as the command returns.


Hmm, that has not been my experience with 3.7.6 and 3.7.8.   Perhaps there is a question of semantics here or definition of what the command precisely does.

What I found is that the command returns fairly quickly, in a few seconds.  In that amount of time, the old brick is removed from the volume configuration and the new brick is added to the volume configuration.  As soon as the replace-brick command is done, the "volume info" will show the new configuration with the old brick gone and the new brick included.  So in the sense of volume configuration, it is complete.

But the data is not moved or healed at this point; that is only starting.  The heal process will then proceed separately after the "replace-brick" command.


I certainly think of the overall brick replacement process as including the full replication of data to the new brick, even if the "replace-brick" command does not do that.  I imagine other people might think the same way also.  A new empty brick isn't protecting your replicated data, so it is an incomplete replacement.  


Older documentation certainly refers to "replace brick start".  I couldn't find any 3.7 documentation that explained why that was gone, and why "commit force" was the only option available now.  I just got errors at the command line trying to do "start".  I think it would help if the new documentation was a little clearer about this, and how to look at heal info to find out when your brick replacement is fully finished.

That's my opinion :-)


- Alan