-
Notifications
You must be signed in to change notification settings - Fork 31
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[#59] Include row level deletion in tombstone count #60
Conversation
@jwiencek3 would you be willing to give this a try before I merge and make a release? I'm hoping this will fix the issue you reported and just wanted to make sure that was the case before doing a release. |
testing right now. an you explain this section of the output: Is the "Count" how many tombstones were removed when a minor compaction occurred? Estimated tombstone drop times: |
That's correct. That's printing a histogram that is the estimated distribution of tombstones would be dropped by compactions happening after that time. EDIT: Need to look into that more to verify that, I had always thought that was the case, but i may be wrong. |
It looks like there was a bug in cassandra (CASSANDRA-12208) where the calculation for estimated droppable tombstones could be incorrect:
further..
This is why I was confused. It appears the times in the histogram don't account for gc_grace_seconds which is why it looks like the time at which data was deleted, not when it would be cleaned up. The issue was fixed in 3.0.9+ and 3.10+, but i'm not sure how it factors into the histogram of the distribution of data since this was mentioning the "Estimated droppable tombstones" metric. Since this histogram is coming from C* itself it's not something I think we can easily control in sstable-tools. I can look into this later (tonight) as a C* issue itself. |
Thinking about it more, we could sort of cheat and add gc_grace_seconds to the timestamps shown in the histogram, then the values would look right. I see that's basically what the patch for that issue does (adds a |
Per #59, it appears that row level deletions are not considered in tombstone count. This fixes this issue.