You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Ctest (Jira)" <ji...@apache.org> on 2020/02/16 23:00:00 UTC
[jira] [Updated] (HBASE-23702) `hbase.hstore.flusher.count` setting
to 0 breaks HMaster
[ https://issues.apache.org/jira/browse/HBASE-23702?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Ctest updated HBASE-23702:
--------------------------
Attachment: HBASE-23702-000.patch
> `hbase.hstore.flusher.count` setting to 0 breaks HMaster
> --------------------------------------------------------
>
> Key: HBASE-23702
> URL: https://issues.apache.org/jira/browse/HBASE-23702
> Project: HBase
> Issue Type: Bug
> Components: regionserver
> Affects Versions: 2.2.3
> Environment: HBase-2.2.3
> Java openjdk 1.8.0_222
> Ubuntu 18.04
> Reporter: Ctest
> Priority: Major
> Attachments: HBASE-23702-000.patch, HBASE-23702-000.patch
>
>
> Summary
> The configuration parameter `hbase.hstore.flusher.count` allows `0` (which turns out to be a bad value after I understand the source code).
> When we set `hbase.hstore.flusher.count` to 0, HMaster will fail in 200,000 ms without any error message, but
> ```
> java.lang.RuntimeException: Master not initialized after 200000ms
> ```
>
> There’s no messages pinpointing `hbase.hstore.flusher.count` being the root cause.
>
> 0 means `MemStoreFlusher` is not alive. During the initialization time of HMaster, `isHealthy()` will be called and it will try to check if `MemStoreFlusher` is alive or not. However, since we set `hbase.hstore.flusher.count` to 0, there will be no flusher created at the first place, which cause`isHealthy()` to return false. As a consequence, HRegionServer does not enter its main loop.
>
> *How To Reproduce*
>
> The version of HBase: 2.2.3
>
> # Set the value of configuration parameter `hbase.hstore.flusher.count` to `0` in “conf/hbase-site.xml”
> # Start hbase by running “bin/start-hbase.sh”
> # Wait for 3-4 minutes
> # Open the log file “logs/hbase-<username>- master -<pcname>.log”
--
This message was sent by Atlassian Jira
(v8.3.4#803005)