[
https://issues.apache.org/jira/browse/HIVE-11957?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15431422#comment-15431422
]
Eugene Koifman commented on HIVE-11957:
---------------------------------------
Adding queryID to a txn doesn't make sense - for example, in multi-statement
txns there will be multiple queries that are part of a txn.
Filtering by DB/table, etc would require a join of TXNS to HIVE_LOCKS and would
produce unstable results as queries acquire/release locks (multi-statement txn
case). So we best leave this alone.
Filtering by AgentId seems useful but lower priority - let's wait until someone
asks for it.
Adding a SessionID may be more useful in HS2 type environments.
For Streaming ingest case agentID that is passed in by the caller makes sense
for both locks and transactions.
We are putting different data into the same field here, but we already include
user and hostname so it should be sufficient to disambiguate.
We should definitely include TXN_STARTED/TXN_LAST_HEARTBEAT timestamps in the
SHOW TRANSACTIONS output.
> SHOW TRANSACTIONS should show queryID/agent id of the creator
> -------------------------------------------------------------
>
> Key: HIVE-11957
> URL: https://issues.apache.org/jira/browse/HIVE-11957
> Project: Hive
> Issue Type: Improvement
> Components: Transactions
> Affects Versions: 1.0.0
> Reporter: Eugene Koifman
> Assignee: Wei Zheng
>
> this would be very useful for debugging
> should also include heartbeat/create timestamps
> would be nice to support some filtering/sorting options, like sort by create
> time, agent id. filter by table, database, etc
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
