seize: enable support for frozen containers #2514
Merged
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
rst0git
force-pushed
the
2024-11-04-seize-checkpointing-freezen-containers
branch
from
70b3ad9
to
511d073
Compare
avagin
reviewed
Nov 5, 2024
|
Doesn't this break the expectations of the container engines. You wrote they freeze the container to avoid changes to the container file-system. Does the container now continue to run with your change? |
rst0git
force-pushed
the
2024-11-04-seize-checkpointing-freezen-containers
branch
2 times, most recently
from
d39fc15
to
979e277
Compare
avagin
reviewed
Nov 7, 2024
Strictly speaking, this expectation was not right even before this change. CRIU does file system changes while dumping processes. For example, it creates ghost files. I think the right expectation here is that file systems are not changed after dumping processes and this statement isn't affected by this change. |
|
LGTM. Thanks. |
Container runtimes like CRI-O and containerd utilize the freezer cgroup to create a consistent snapshot of container root filesystem (rootfs) changes. In this case, the container is frozen before invoking CRIU. After CRIU successfully completes, a copy of the container rootfs diff is saved, and the container is then unfrozen. However, the `cuda-checkpoint` tool is not able to perform a 'lock' action on frozen threads. To support GPU checkpointing with these container runtimes, we need to unfreeze the cgroup and return it to its original state once the checkpointing is complete. To reflect this new behavior, the following changes are applied: - `dont_use_freeze_cgroup(void)` -> `set_compel_interrupt_only_mode(void)` - `bool freeze_cgroup_disabled` -> `bool compel_interrupt_only_mode` - `check_freezer_cgroup(void)` -> `prepare_freezer_for_interrupt_only_mode(void)` Note that when `compel_interrupt_only_mode` is set to `true`, `compel_interrupt_task()` is used instead of `freeze_processes()` to prevent tasks from running during `criu dump`. Fixes: checkpoint-restore#2508 Signed-off-by: Radostin Stoyanov <[email protected]>
Signed-off-by: Radostin Stoyanov <[email protected]>
rst0git
force-pushed
the
2024-11-04-seize-checkpointing-freezen-containers
branch
from
979e277
to
495e39e
Compare
rst0git
deleted the
2024-11-04-seize-checkpointing-freezen-containers
branch
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.
Container runtimes like CRI-O and containerd utilize the freezer cgroup to create a consistent snapshot of container rootfs changes. In this case, the container is frozen before invoking CRIU. Once CRIU successfully completes, a copy of the container rootfs diff is saved, and then the container is unfrozen. To enable GPU checkpointing support with these runtimes, we need to unfreeze the cgroup and restore it to its original state at the end.
When the CUDA plugin is installed, container checkpointing with Kubernetes fails, even for containers that don't use GPUs. This patch aims to resolve this issue.
Fixes: #2508