[
https://issues.apache.org/jira/browse/CARBONDATA-4098?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yahui Liu updated CARBONDATA-4098:
----------------------------------
Description:
CARBONDATA-4088 introduce the expiring map to expire the carbon table cache by
time. But it will have one problem: during query, if cache is being expired,
queries on the table may fail with NullPointerException.
If users don't want their queries to be failed, they need to choose a proper
value for the configuration carbon.metacache.expiration.seconds: at the time of
cache expired, the table will never be queried. For example, customer will have
a new table every day, and this table will be queried only at that day. So he
can choose 1 week as the property value so that this table cache will only leak
for one week and after one week will be cleared.
So mostly this nullpointerexception will not happen if user chooses the proper
value of carbon.metacache.expiration.seconds. But still we will have one Jira
to track this seldom behavior and maybe in future will fix it.
was:CARBONDATA-4088 introduce the expiring map to expire the stale carbon
table cache. But it will have a problem: during query, if
> NullPointerException will be thrown during query if at the time carbon table
> cache is being expired
> ---------------------------------------------------------------------------------------------------
>
> Key: CARBONDATA-4098
> URL: https://issues.apache.org/jira/browse/CARBONDATA-4098
> Project: CarbonData
> Issue Type: Bug
> Components: sql
> Affects Versions: 2.1.0
> Reporter: Yahui Liu
> Priority: Major
>
> CARBONDATA-4088 introduce the expiring map to expire the carbon table cache
> by time. But it will have one problem: during query, if cache is being
> expired, queries on the table may fail with NullPointerException.
> If users don't want their queries to be failed, they need to choose a proper
> value for the configuration carbon.metacache.expiration.seconds: at the time
> of cache expired, the table will never be queried. For example, customer will
> have a new table every day, and this table will be queried only at that day.
> So he can choose 1 week as the property value so that this table cache will
> only leak for one week and after one week will be cleared.
> So mostly this nullpointerexception will not happen if user chooses the
> proper value of carbon.metacache.expiration.seconds. But still we will have
> one Jira to track this seldom behavior and maybe in future will fix it.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
