What I really meant is trying to get cluster status directly through ZK
API. Your approach a bit different from what I meant but it's a nice one as
it seems will work across versions 4 and 5.
On Thursday, April 9, 2015, Shalin Shekhar Mangar > wrote:
> I don't quite follow. Are you saying that yo
I don't quite follow. Are you saying that you intend to use the ZK REST API
to fetch live_nodes and then send the 'clusterstatus' API call to one of
the live nodes?
On Thu, Apr 9, 2015 at 7:13 PM, Ahmed Adel wrote:
> In fact, the advantage I see of using ZK is that we don't have to iterate
> ove
In fact, the advantage I see of using ZK is that we don't have to iterate
over nodes in case the first node receiving that request is down, whereas,
by using ZK REST API, we can do that in a single request as I assume we can
check live_nodes (in case this approach is guaranteed when using Solr 4.x)
Yes, you can use the 'clusterstatus' API which will return an aggregation
of all states. See
https://cwiki.apache.org/confluence/display/solr/Collections+API#CollectionsAPI-api18
On Thu, Apr 9, 2015 at 5:52 PM, Ahmed Adel wrote:
> Hi Shalin,
>
> Thanks for your response. I'm actually looking in
Hi Shalin,
Thanks for your response. I'm actually looking inside ZooKeeper in order to
obtain highest availability. What I expected is that clusterstate.json
contains the aggregation of all state.json children nodes of each
collection. But your second paragraph explains the behavior I see in Solr
Hi Ahmed,
Can you give more details? What did you expect and what was the actual?
Also, are you looking directly at the clusterstate.json inside ZooKeeper or
are you using the 'clusterstatus' Collection API?
You shouldn't look at the clusterstate.json directly because 1) things like
live-ness is
Hi All,
On Solr 5.0 and ZK 3.4.6 sometimes clusterstate.json does not reflect the
aggregation of states of collections, the latter is always correct. I could
verify this from the admin panel (under Tree view) and from ZKCli. Is there
something I'm missing that could generate this issue?
--
A. A
