[
https://issues.apache.org/jira/browse/IMPALA-5105?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Bikramjeet Vig closed IMPALA-5105.
----------------------------------
Resolution: Won't Fix
Behavior has changed, impalad works off of local catalog cache. For a new query
that requires metadata not in the cache, the query fails with a descriptive
message "Error making an RPC call to Catalog server"
> Improve error message in the event catalogd crashes
> ---------------------------------------------------
>
> Key: IMPALA-5105
> URL: https://issues.apache.org/jira/browse/IMPALA-5105
> Project: IMPALA
> Issue Type: Improvement
> Components: Frontend
> Reporter: David Knupp
> Priority: Major
>
> IMPALA-4704 documents a case whereby impalad opens up ports 21000 and 21050
> before the catalog update has been received, and this can lead to queries
> failing with the message:
> {noformat}
> AnalysisException: This Impala daemon is not ready to accept user requests.
> Status: Waiting for catalog update from the StateStore.
> {noformat}
> Also, we have seen cases where queries fail when the catalogd crashes for
> some reason or another, and the exact same error is produced:
> {noformat}
> [impalad-2.abc.xyz.com:21000] > show tables;
> Query: show tables
> ERROR: AnalysisException: This Impala daemon is not ready to accept user
> requests. Status: Waiting for catalog update from the StateStore.
> {noformat}
> One error is a recoverable timing issue, the other is a crashed processed. It
> would be nice if these two conditions were differentiated by unique error
> messages.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
