In general, you can't. that's to ensure not all the replicas are offline at
once.

However, vos release has the -stayonline switch, which will do them all in
parallel...
it changes the order of how refreshed replicas come online,
instead of a small number at a time until they are done, all the data is
moved, and then the replicas
all come online at the same time.



On Wed, Sep 18, 2013 at 4:13 AM, Antony Mayi <[email protected]> wrote:

> Hi,
>
> When releasing a volume I can see it starting an operation called
> ForwardMulti. I have 5 remote replicas for a volume and it seems to start
> three ForwardMulti sessions (two with single replica fileserver (ro3, ro5)
> and 1 with three (ro1 + ro2 + ro4)). Based on the observation (see the
> release report at the bottom) the sessions are started in sequence and all
> replications within same session run in parallel. If each individual
> replication operation takes T_repl time it means each ForwardMulti session
> also takes T_repl time and in my case the whole release operation takes
> 3*T_repl time. That makes me to think that it could have taken only
> 1*T_repl time had all the replications been made within single ForwardMulti
> session.
>
> I don't seem to be able to find any information about what the ForwardMulti
> is and why some replications are made within single ForwardMulti session
> and others in separate ones. Just to give some background all the roX
> fileservers are in same rack, on same network.
>
> How can I make all the replication to run within single ForwardMulti
> session?
>
> Thanks,
> Antony.
>
>
> The typical release in my case produces following report:
>
> /usr/sbin/vos release proj.prj1 -verbose
> proj.prj1
>     RWrite: 536871449     ROnly: 536871450     Backup: 536871451
>     number of sites -> 7
>        server rw1 partition /vicepa RW Site
>        server rw1 partition /vicepa RO Site
>        server ro5 partition /vicepa RO Site
>        server ro1 partition /vicepa RO Site
>        server ro4 partition /vicepa RO Site
>        server ro2 partition /vicepa RO Site
>        server ro3 partition /vicepa RO Site
> This is a complete release of volume 536871449
> Re-cloning permanent RO volume 536871450 ... done
> Getting status of parent volume 536871449... done
> Starting transaction on RO clone volume 536871450... done
> Setting volume flags for volume 536871450... done
> Ending transaction on volume 536871450... done
> Replacing VLDB entry for proj.prj1... done
> Starting transaction on cloned volume 536871450... done
> Updating existing ro volume 536871450 on ro5 ...
> Starting ForwardMulti from 536871450 to 536871450 on ro5 (as of Wed Sep 18
> 08:23:33 2013).
> Updating existing ro volume 536871450 on ro1 ...
> Updating existing ro volume 536871450 on ro4 ...
> Updating existing ro volume 536871450 on ro2 ...
> Starting ForwardMulti from 536871450 to 536871450 on ro1 and ro4 and ro2
> (as of Wed Sep 18 08:23:33 2013).
> Updating existing ro volume 536871450 on ro3 ...
> Starting ForwardMulti from 536871450 to 536871450 on ro3 (as of Wed Sep 18
> 08:23:33 2013).
> updating VLDB ... done
> Released volume proj.prj1 successfully
>
>


-- 
Derrick

Reply via email to