David Smiley created SOLR-11877:
-----------------------------------
Summary: DocCollection.getStateFormat is buggy; sometimes wrong
Key: SOLR-11877
URL: https://issues.apache.org/jira/browse/SOLR-11877
Project: Solr
Issue Type: Bug
Security Level: Public (Default Security Level. Issues are Public)
Components: SolrCloud
Reporter: David Smiley
The ability to know what the collection / ClusterState "stateFormat" integer is
per collection is a bit buggy. After looking at the problem for awhile, I
think it would help a lot if DocCollection accepted the stateFormat param to
thus force the caller to consider it. Today the caller passes the znode
path... but in several places we pass ZkStateReader.CLUSTER_STATE or null (both
implying stateFormat==1) which is sometimes not accurate. Nobody seems to
actually need the znode path from the DocCollection but the stateFormat is
requested in a bunch of places.
The particular bug I see is
org.apache.solr.client.solrj.impl.HttpClusterStateProvider#fetchClusterState
passing ZkStateReader.CLUSTER_STATE when in reality it has no clue what the
format is because our cluster_status API doesn't return the stateFormat.
Maybe getStateFormat should return an Integer object and hence can be null if
it's not known.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
