-
Notifications
You must be signed in to change notification settings - Fork 1.1k
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
SEGV after failing to map contiguous pages by MmapAllocator #9135
Comments
FYI @majetideepak |
A different stack trace I see:
|
Yes, I think ultimately it comes from the same source. The stack to lead up to it from
The |
After we handle the
|
PR resolved the issue. |
Summary: This PR is addressing this issue facebookincubator#9135 Pull Request resolved: facebookincubator#9228 Reviewed By: Yuhta Differential Revision: D55609322 Pulled By: xiaoxmeng fbshipit-source-id: c1921fd5551c553384fde4ffddf80ad659471094
Bug description
Under certain conditions memory cannot be allocated by the MemoryAllocator. This occasionally on clusters running TPCDS queries.
Prestissimo does not explicitly set the option
useMmapArena
and the default value offalse
is used by Velox. As a result, contiguous memory requests use themmap
system call.The code in
allocateContiguousImpl
performing the allocation may callmmap
.If the
mmap
call fails then data will be set toMAP_FAILED (-1)
anderrno
is set.The code has a TODO comment to handle MAP_FAILED which == -1.
The proposed solution to prevent the SEGV and return the error (fail query with insufficient resources) would be to check for MAP_FAILED
System information
Host with 122 GiB memory available. Async data cache enabled by default.
system-memory-gb=108
query-memory-gb=102
query.max-memory-per-node=102GB
memory-arbitrator-kind=SHARED
task.max-drivers-per-task=16
Relevant logs
The text was updated successfully, but these errors were encountered: