> On Nov. 29, 2017, 4:50 p.m., Sergio Pena wrote:
> > sentry-provider/sentry-provider-db/src/main/java/org/apache/sentry/service/thrift/HMSFollower.java
> > Lines 251-252 (patched)
> > <https://reviews.apache.org/r/63881/diff/2/?file=1901606#file1901606line251>
> >
> >     This line is making me some noise. If a user upgrades  their cluster, 
> > then the MSentryHmsNotification table will be empty initially; and if the 
> > user does not execute any DDL command for a long time, then this message 
> > will appear every 500ms on the log.
> >     
> >     Is this log message necessary here?
> 
> Arjun Mishra wrote:
>     Yeah it can be debug. I'll change it
> 
> Na Li wrote:
>     Sergio, in this case, sentry will keep on getting full snapshot when 
> upser upgrade their cluster, and not execute any DDL command for a long time. 
> It is not desirable. We should re-visit this logic on when to get full 
> snapshot.

Agree. I think we should write a snapshot even if it is empty.

This below code is the one returning if an attempt to create a snapshot doesn't 
do anything. We may need to persist the snapshot ID at least if it is empty.

private long createFullSnapshot() {
    PathsImage snapshotInfo = client.getFullSnapshot();
    if (snapshotInfo.getPathImage().isEmpty()) {
       return snapshotInfo.getId();
    }
}


- Sergio


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/63881/#review192133
-----------------------------------------------------------


On Nov. 29, 2017, 5:43 p.m., Arjun Mishra wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/63881/
> -----------------------------------------------------------
> 
> (Updated Nov. 29, 2017, 5:43 p.m.)
> 
> 
> Review request for sentry, kalyan kumar kalvagadda, Na Li, Sergio Pena, and 
> Vadim Spector.
> 
> 
> Repository: sentry
> 
> 
> Description
> -------
> 
> Currently we only seem to log when all conditions that lead to getting a 
> snapshot are validated, and eventually a full snapshot is received. However, 
> if for some reason a particular condition was invalid we don't log as to why. 
> This leaves no room for knowing definitely what might have been the reason as 
> to why a snapshot was not received from HMS.
> 
> 
> Diffs
> -----
> 
>   
> sentry-provider/sentry-provider-db/src/main/java/org/apache/sentry/provider/db/service/persistent/SentryStore.java
>  f32a745ed 
>   
> sentry-provider/sentry-provider-db/src/main/java/org/apache/sentry/service/thrift/HMSFollower.java
>  c1471d118 
>   
> sentry-provider/sentry-provider-db/src/main/java/org/apache/sentry/service/thrift/SentryHMSClient.java
>  eccb40fb6 
> 
> 
> Diff: https://reviews.apache.org/r/63881/diff/5/
> 
> 
> Testing
> -------
> 
> mvn -f sentry-provider/sentry-provider-db/pom.xml test
> 
> 
> Thanks,
> 
> Arjun Mishra
> 
>

Reply via email to