Skip to content
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.

Sign up for GitHub

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

Jump to bottom

Do not cache system.large_* tables #8874

Closed
bhalevy opened this issue Jun 17, 2021 · 2 comments
Closed

Do not cache system.large_* tables #8874

bhalevy opened this issue Jun 17, 2021 · 2 comments
Assignees
@cmm
Labels
area/row cache scale/large partitions
Milestone
4.6

Comments

@bhalevy
Copy link
Member

bhalevy commented Jun 17, 2021

This issue follows up on #7668 (comment)

@slivne said:

in the case a customer has large partitions - we are going to create a real issue for him (and us) with generating range tombstones that can't be evicted out of the cache and only accumulate - those range tombstones are going to keep pilling up since a compaction of such a table ends up usually with another table that has a large partition as well ...

I think we need to mark the system_large* tables as tables that should not be cached to make sure we do not cause an issue and once we have range tombstones handling made better (and we are working on that) consider removing that.
@bhalevy bhalevy mentioned this issue Jun 17, 2021
Closed
@bhalevy
Copy link
Member Author

bhalevy commented Jun 23, 2021

More about the motivation for this issue:

We generate much less range tombstones with
#7668 fixed.
But we still generate range tombstones for sstables
that do hold large_ data.

They accumulate in the cache as per:
#7750

and they need to be evicted gradually, as per:
#3288

So this is a workaround until the above is fixed.
Please add a FIXME comment to undo it when the root cause is fixed.

lauranovich pushed a commit to lauranovich/scylla that referenced this issue Jul 29, 2021
@cmm @lauranovich
disable caching of the system.large_* tables
d142034 
The cache of system.large_{partition,rows,cells} accumulates range
tombstones (scylladb#7750), and those
range tombstones can be evicted only together with their partition
(scylladb#3288).

Making the system.large_* tables uncached should work around the
problem until scylladb#3288 is fixed.

Fixes scylladb#8874
Refs scylladb#7750
Refs scylladb#3288

Signed-off-by: Michael Livshin <michael.livshin@scylladb.com>
Message-Id: <20210623171932.8837-1-michael.livshin@scylladb.com>
@avikivity
Copy link
Member

Not backporting, since #7668 (which was backported) is likely sufficient.

@DoronArazii DoronArazii added this to the 4.6 milestone May 7, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@cmm cmm

Labels
area/row cache scale/large partitions
Projects
None yet
Milestone
4.6
Development

No branches or pull requests
5 participants
@cmm @avikivity @bhalevy @scylladb-promoter @DoronArazii