[ 
https://issues.apache.org/jira/browse/CASSANDRA-15882?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17141928#comment-17141928
 ] 

Alex Petrov commented on CASSANDRA-15882:
-----------------------------------------

[~thatran] tombstone drop time seems to be consistent with the tombstone 
expiration: histogram is built using local delete time. It also seems logical 
that a tombstone will hang around in the sstable starting when it shouldn't be 
visible anymore + gc grace. I might be missing something, so I'd appreciate if 
you elaborate.

> sstablemetadata should show tombstone timestamp not just the local drop time
> ----------------------------------------------------------------------------
>
>                 Key: CASSANDRA-15882
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-15882
>             Project: Cassandra
>          Issue Type: Improvement
>            Reporter: Thanh
>            Priority: Normal
>
> issue/request:
> sstablemetadata should show tombstone timestamp not just the local drop time
>  
> I'm  not sure what all versions this exists in but the sstablemetadata 
> tombstone "drop time" only just shows the local delete time of the 
> tombstones.  What's more important (what the tombstone expiration time 
> depends on) is the tombstone writetime (tombstone timestamp).  It's possible 
> to, and a surprising number of users have done this, write a tombstone with a 
> timetamp in the future (by doing "DELETE...USING TIMESTAMP 
> <future_timestamp>").  When this happens, the tombstone hangs around in 
> sstables for a lot longer than users expect because
> {code:java}
> tombstone_expiration_time = gc_grace_seconds + tombstone_timestamp{code}
> NOT
> {code:java}
>  tombstone_expiration_time = gc_grace_seconds + 
> tombstone_local_delete_time{code}
> and the sstablemetadata output doesn't help here because it only shows the 
> local delete time, not the actual tombstone timestamp that's used to 
> calculate tombstone expiration time.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org

Reply via email to