phoenix-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Samarth Jain (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (PHOENIX-3947) Increase scan time out for partial index rebuild and retry only once
Date Fri, 07 Jul 2017 00:44:00 GMT

     [ https://issues.apache.org/jira/browse/PHOENIX-3947?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Samarth Jain updated PHOENIX-3947:
----------------------------------
    Attachment: PHOENIX-3947.patch

[~jamestaylor], please review. 

I am glad I wrote the test as I ended up discovering bug in our existing QueryUtil.getConnectionOnServer()
code. It turned out the code in PropertiesUtil#extractProperties wasn't setting the properties
correctly.

Before PropertiesUtil#extractProperties:
{code}
            while (iterator.hasNext()) {
                Map.Entry<String, String> entry = iterator.next();
                props.setProperty(entry.getKey(), entry.getValue());
            }
{code}

Now PropertiesUtil#extractProperties:
{code}
            while (iterator.hasNext()) {
                Map.Entry<String, String> entry = iterator.next();
                // set the property from config only if props doesn't have it already
                if (props.getProperty(entry.getKey()) == null) {
                    props.setProperty(entry.getKey(), entry.getValue());
                }
            }
{code}

> Increase scan time out for partial index rebuild and retry only once
> --------------------------------------------------------------------
>
>                 Key: PHOENIX-3947
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-3947
>             Project: Phoenix
>          Issue Type: Bug
>            Reporter: James Taylor
>            Assignee: Samarth Jain
>              Labels: globalMutableSecondaryIndex
>         Attachments: PHOENIX-3947.patch
>
>
> The scan done from MetaDataRegionObserver needs to have a higher timeout so that it can
always complete successfully. The retries should be limited to one too. Upon failure, we should
disable the index and set the INDEX_DISABLE_TIMESTAMP to zero to prevent further attempts
to rebuild the index (as a failure would be an indication that something is awry and manual
intervention should be required to rebuild the index completely).



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Mime
View raw message