I've recently stood up a SolrCloud 7.7.1 cluster on AWS EC2 instances, with a
dedicated zookeeper ensemble (3.4.13). It looks to have the basic trappings
of a 'test' collection, I can access the Solr admin UI through an SSH tunnel
as well as run the following API calls with success (from a solr node in the
cluster):

curl http://localhost:8983/solr/test/admin/ping
{
  "responseHeader":{
    "zkConnected":true,
    "status":0,
    "QTime":0,
    "params":{
      "q":"{!lucene}*:*",
      "distrib":"false",
      "df":"_text_",
      "rows":"10",
      "echoParams":"all"}},
  "status":"OK"}

curl http://localhost:8983/solr/test/select?q=*:*
{
  "responseHeader":{
    "zkConnected":true,
    "status":0,
    "QTime":10,
    "params":{
      "q":"*:*"}},
  "response":{"numFound":0,"start":0,"maxScore":0.0,"docs":[]
  }}

However, attempting the same API calls from a remote client results in
connection resets. It does not appear to be a firewall issue, as neither
netcat nor SSH have issue. The connection is being made over private IP
addresses within the network, from the same machine I use to SSH into a Solr
EC2 instance (10.131.200.233 as an example).

nc -vz 10.131.200.233 8983
found 0 associations
found 1 connections:
     1: flags=82<CONNECTED,PREFERRED>
        outif gpd0
        src 172.16.253.5 port 50830
        dst 10.131.200.233 port 8983
        rank info not available
        TCP aux info available

Connection to 10.131.200.233 port 8983 [tcp/*] succeeded!

curl http://10.131.200.233:8983/solr/test/admin/ping
curl: (56) Recv failure: Connection reset by peer
curl http://10.131.200.233:8983/solr/test/select?q=*:*
curl: (56) Recv failure: Connection reset by peer

I am unable to access the Solr Admin UI without an SSH tunnel, contrary to
the doc at
https://lucene.apache.org/solr/guide/7_7/aws-solrcloud-tutorial.html. My
internet searches have resulted in a fair bit of confusion, but it seems
like Solr is denying anything not localhost as a security feature. In our
use case we have a number of client applications already deployed on a fleet
of other EC2 instances and would like to give them API search capabilities
against this up-and-coming SolrCloud cluster. I was thinking just to put an
AWS ALB/ELB in front of the Solr nodes, but the primary concern is simply
getting remote queries working in the first place.




--
Sent from: http://lucene.472066.n3.nabble.com/Solr-User-f472068.html

Reply via email to