[
https://issues.apache.org/jira/browse/HIVE-14765?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15538250#comment-15538250
]
Barna Zsombor Klara edited comment on HIVE-14765 at 10/3/16 11:56 AM:
----------------------------------------------------------------------
Hi [~sershe]
I think I found a possible reason for this. Would you mind if I take a look at
it?
Edit: Unfortunately my theory doesn't hold water.... I thought it might be
because we aren't clearing the ConcurrentHashMap in the CodahaleMetrics used to
store the Guages in (besides the metrics registry). But it turns out the map is
completely superfluous, we aren't using it at all. Can you give me any details
as to when you could see these warnings in the logs? Was the HMS running as a
service or in embedded mode? I think we could run into it with an embedded HMS
if the HMSClient is recreated.
Thanks,
Zs.
was (Author: zsombor.klara):
Hi [~sershe]
I think I found a possible reason for this. Would you mind if I take a look at
it?
> metrics - gauge overwritten messages
> ------------------------------------
>
> Key: HIVE-14765
> URL: https://issues.apache.org/jira/browse/HIVE-14765
> Project: Hive
> Issue Type: Bug
> Reporter: Sergey Shelukhin
>
> {noformat}
> 2016-09-14T21:09:55,553 WARN [HiveServer2-HttpHandler-Pool: Thread-48]:
> metrics2.CodahaleMetrics (CodahaleMetrics.java:addGauge(304)) - A Gauge with
> name [init_total_count_dbs] already exists. The old gauge will be
> overwritten, but this is not recommended
> 2016-09-14T21:09:55,553 WARN [HiveServer2-HttpHandler-Pool: Thread-48]:
> metrics2.CodahaleMetrics (CodahaleMetrics.java:addGauge(304)) - A Gauge with
> name [init_total_count_tables] already exists. The old gauge will be
> overwritten, but this is not recommended
> 2016-09-14T21:09:55,554 WARN [HiveServer2-HttpHandler-Pool: Thread-48]:
> metrics2.CodahaleMetrics (CodahaleMetrics.java:addGauge(304)) - A Gauge with
> name [init_total_count_partitions] already exists. The old gauge will be
> overwritten, but this is not recommended
> {noformat}
> Might have something to do with metastore being a threadlocal (just shooting
> in the dark)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
