0

We are using Solr 4.3.0. We have a solr cloud setup with two servers, each collection has 2 shard and 2 replicas. Within that 2 shared server run in UK DC and 2 replica run in AUS DC using cross Data centre. Now We are add/update document it will take time to process. If We shut down the 2 replica of AUS DC(Only 2 shared server runs in UK DC) it works fine means process completed faster. We think it could be delayed due to a network.

Here problem question is, Is there any way to send add/update document request directly to only shared server first and it will replicate the shared document with replica with asynchronous call.

user3326695
  • 1
  • You should be able to target the shard leaders only without waiting for the change to propogate to the shard replicas. If you are experiencing latency on add/update due to propogation, I would presume it is because you are committing the changes. Have you tried using autocommit on the server (either by number of docs or time) and *not* explicitly committing your adds/updates? That should batch up the changes asynchronously and allow your add/update process to complete quickly – gary schulte Feb 19 '14 at 22:35

1 Answers1

0

Answer to your question is yes. you can specify static ips of your shared server in solrj client. But in that case you have no HA guarantee, as the shared node goes down, you are out of business. And you really dont need to do that cause solrcloud manages this internally, as it uses zookeeper to find out leaders in the cluster and forwards writes to the leaders only.Now these leaders triggers sync process for replicas by replaying transaction log since last commit.