zookeeper-bookkeeper-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From John Nagro <jna...@hubspot.com>
Subject Re: rolling restarts
Date Thu, 03 May 2012 15:23:03 GMT
Thanks Ivan!

On Thu, May 3, 2012 at 11:20 AM, Ivan Kelly <ivank@apache.org> wrote:

> On Thu, May 03, 2012 at 10:45:51AM -0400, John Nagro wrote:
> > Okay - and doing the upgrade to a live cluster? I can do that to 1
> machine
> > at a time while the cluster is still online and everything should operate
> > as normal? I have 5 machines, use 3 at a time, quorum of 2.
> In terms of rolling restarts, if ensemble size if 3 and you have 5
> machines, you can bring down one machine at a time and it'll be
> fine. Any clients writing to that machine will move very quickly to
> using one of the other machines (it'll see the network drop, and
> rebuild an ensemble from the other machines).
>
> Regarding whether you need to run the upgrade code, it depends on what
> the old version of bookkeeper you are using is. If you need to run
> upgrade, the bookie will warn you and fail to start.
>
> https://issues.apache.org/jira/browse/BOOKKEEPER-163 is actually the
> change that requires the upgrade framework.
>
> -Ivan
>

Mime
View raw message