[ https://issues.apache.org/jira/browse/PHOENIX-1072?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14059177#comment-14059177 ]
Jeffrey Zhong commented on PHOENIX-1072: ---------------------------------------- {quote} if the client can't reach the master, does the error message still show up {quote} Yes, error message will still show up like norm. The child process will be killed only when parent process(sqlline.py) dies by ctrl-c in all other cases there is no action. This issue before is that even when sqlline can't connect to Phoenix and sqlline.py is killed by ctrl-c, the sqlline java child process still runs for quite a while and have to manually kill it. {quote} It'd be nice if hbase had the ability to specify retry info when making a connection, as a parameter rather than just globally through the Config. If you agree, maybe you could file a JIRA {quote} Do you mena a HBase JIRA? Currently we could so similar things by passing a temporary conf setting & restore it back(not elegant though) but there are issues: 1) the created connection(with different retries) might be cached by HBase and reused later 2) test cases needs long wait for underlying hbase test cluster is ready. There are some issues inside HBase getMaster which should do smart retries and fail fast when zookeeper isn't available or parent znode is wrong etc. Let me file a bug on that. > Fast fail sqlline.py when pass wrong quorum string or hbase cluster hasnt' > started yet > --------------------------------------------------------------------------------------- > > Key: PHOENIX-1072 > URL: https://issues.apache.org/jira/browse/PHOENIX-1072 > Project: Phoenix > Issue Type: Improvement > Reporter: Jeffrey Zhong > Assignee: Jeffrey Zhong > Fix For: 5.0.0, 3.1, 4.1 > > Attachments: phoenix-1072-v1.patch, phoenix-1072.patch > > > Currently sqlline.py will retry 35 times to talk to HBase master when the > passed in quorum string is wrong or the underlying HBase isn't running. > In that situation, Sqlline will stuck there forever. The JIRA is aiming to > fast fail sqlline.py. -- This message was sent by Atlassian JIRA (v6.2#6252)