smp: disable mlock on overcommit #2575
Open
+7
−4
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
xemul
reviewed
Dec 24, 2024
| @@ -3814,7 +3814,7 @@ smp_options::smp_options(program_options::option_group* parent_group) | |||
| , memory(*this, "memory", std::nullopt, "memory to use, in bytes (ex: 4G) (default: all)") | |||
| , reserve_memory(*this, "reserve-memory", {}, "memory reserved to OS (if --memory not specified)") | |||
| , hugepages(*this, "hugepages", {}, "path to accessible hugetlbfs mount (typically /dev/hugepages/something)") | |||
| , lock_memory(*this, "lock-memory", {}, "lock all memory (prevents swapping)") | |||
| , lock_memory(*this, "lock-memory", {}, "lock all memory (prevents swapping). Disable for overprovisioned environments") | |||
There was a problem hiding this comment.
The help text for --overprovisioned should be updated too
Memory locking keeps pages in physical memory, and decreases the amount of pages available for swapping on an overcommit setup. Thus, it can decrease fairness, increase memory pressure, and cause OOMs. So, it is less suitable for a cooperative overcommit setup. This change disables memory locking on overcommit (overprovisioned - I hope I can rename that config). Also, it reuses the mlock variable. I assume that this change shouldn't break reasonable configs.
tomershafir
force-pushed
the
smp-mlock-disable-on-overcommit
branch
from
cc1b6bb to
3e4d3eb
Compare
|
Wasn't mlock default off for all environments? |
avikivity
reviewed
Dec 24, 2024
| @@ -4332,7 +4332,7 @@ void smp::configure(const smp_options& smp_opts, const reactor_options& reactor_ | |||
| hugepages_path = smp_opts.hugepages.get_value(); | |||
| } | |||
| auto mlock = false; | |||
| if (smp_opts.lock_memory) { | |||
| if (smp_opts.lock_memory && !reactor_opts.overprovisioned) { | |||
| mlock = smp_opts.lock_memory.get_value(); | |||
There was a problem hiding this comment.
It's default off, so why this extra work?
If someone specified both --overprovisioned (which really should have been named --overcommitted) and --lock-memory 1, we can assume they know what they're doing. Maybe only the CPU is overcommitted.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Memory locking keeps pages in physical memory, and decreases the amount of pages available for swapping on an overcommit setup. Thus, it can decrease fairness, increase memory pressure, and cause OOMs. So, it is less suitable for a cooperative overcommit setup.
This change disables memory locking on overcommit (overprovisioned - I hope I can rename that config). Also, it reuses the mlock variable. I assume that this change shouldn't break reasonable configs.