Thanks Margus.
1) I don't  do anything (at least explicitly) with zookeeper settings or
the offsets. All I do in the topology, I specify the address of the
zookeeper (ZkHosts). Here is some the lines in between:
2015-01-25T13:26:52.288-0600 s.k.PartitionManager [INFO] Read last commit
offset from zookeeper: 7; old topology_id:
a001b24c-d99c-49f1-8b60-573aac41937e - new topology_id:
ad09f254-ca58-4298-8f57-09741389da56
2015-01-25T13:26:52.288-0600 s.k.PartitionManager [INFO] Starting Kafka
localhost:1 from offset 7
2015-01-25T13:26:52.297-0600 s.k.PartitionManager [INFO] Read partition
information from: /storm/partition_0  --> null
2015-01-25T13:26:52.302-0600 s.k.PartitionManager [INFO] No partition
information found, using configuration to determine offset

2) So /storm actually is an absolute path?

On Sun, Jan 25, 2015 at 2:45 PM, Margus Roo <[email protected]> wrote:

>  Maybe you overwrote it in your topology code?
> zk path in example /storm is a place where storm keeps statuses about
> storm in example.
>
> Margus (margusja) Roohttp://margus.roo.ee
> skype: margusja
> +372 51 480
>
> On 25/01/15 22:16, clay teahouse wrote:
>
> Hi All,
>
>  A couple of questions:
> 1)
> What does storm.zookeeper.root point to? The definition says storm root
> directory location. The default value is /storm. But I am not sure I
> understand what this means. If I have external zookeeper running, what
> should this configuration point to?
>
>  2) I am running my topology in cluster mode, and I said with external
> zookeeper (its configuration is specified in storm.yaml). But I am seeing
> the following lines in my worker logs:
>
>   s.k.DynamicBrokersReader [INFO] Read partition info from zookeeper:
> GlobalPartitionInformation{partitionMap={0=localhost:9092,
> 1=localhost:9092}}
>  ..
> ..
> ..
> s.k.PartitionManager [INFO] No partition information found, using
> configuration to determine offset
>
>  I've read that the second message means that storm is not using the
> external zookeeper. If that is the case, what would be the reason for that.
> thanks,
> Clay
>
>
>

Reply via email to