incubator-blur-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chris Rohr <>
Subject Re: [jira] [Closed] (BLUR-342) Support full cluster operations on random ports
Date Wed, 17 Sep 2014 16:28:10 GMT
Before we had discussed the possiblity of having a configuration that
would state how many shards/controllers etc there are expected to be.
That way we could at least alert if the number drops below expected.
Has that piece been done?

On Wed, Sep 17, 2014 at 12:09 PM, Tim Williams <> wrote:
> they are, mainly to drive startup/shutdown/status scripts.  there's
> still some outstanding questions with respect to a blur-managed
> cluster (as  opposed to a yarn managed one) that still need thought
> through - random ports are but a pit-stop on the journey ahead;)
> --tim
> On Wed, Sep 17, 2014 at 12:05 PM, Chris Rohr <> wrote:
>> So, the console today uses Zookeeper not Thrift for the node
>> information.  So that will have to be redone.  On a related note, are
>> there still going to be files in the blur directory that list all of
>> the nodes (i.e. conf/shards, conf/controllers, etc)?
>> On Wed, Sep 17, 2014 at 11:58 AM, Tim Williams <> wrote:
>>> On Wed, Sep 17, 2014 at 11:54 AM, Chris Rohr <> wrote:
>>>> Tim,
>>>> Quick question on this.  Is this code in the master branch?  If so
>>>> this means that the offline shards and controllers are no longer in
>>>> zookeeper correct?
>>>> Depending on the answers we may need to make some more updates to the
>>>> console prior to releasing 0.2.4
>>> That's right, if you using thrift api nothing should be "broke" in a
>>> real sense.  It'll just never actually report anything as being
>>> offline because shardServerList == onlineShardServers and the same for
>>> controllers. If you still want the console to report 'offline' shards
>>> you'll want to create a local cache with a a configured expiry I
>>> guess.
>>> Thanks,
>>> --tim

View raw message