memcg: introduce non-blocking limit setting option

Setting the max and high limits can trigger synchronous reclaim and/or oom-kill if the usage is higher than the given limit. This behavior is fine for newly created cgroups but it can cause issues for the node controller while setting limits for existing cgroups.

In our production multi-tenant and overcommitted environment, we are seeing priority inversion when the node controller dynamically adjusts the limits of running jobs of different priorities. Based on the system situation, the node controller may reduce the limits of lower priority jobs and increase the limits of higher priority jobs. However we are seeing node controller getting stuck for long period of time while reclaiming from lower priority jobs while setting their limits and also spends a lot of its own CPU.

One of the workaround we are trying is to fork a new process which sets the limit of the lower priority job along with setting an alarm to get itself killed if it get stuck in the reclaim for lower priority job. However we are finding it very unreliable and costly. Either we need a good enough time buffer for the alarm to be delivered after setting limit and potentialy spend a lot of CPU in the reclaim or be unreliable in setting the limit for much shorter but cheaper (less reclaim) alarms.

Let's introduce new limit setting option which does not trigger reclaim and/or oom-kill and let the processes in the target cgroup to trigger reclaim and/or throttling and/or oom-kill in their next charge request. This will make the node controller on multi-tenant overcommitted environment much more reliable.

Signed-off-by Shakeel Butt <[email protected]>

---
Changes since v1:
- Instead of new interfaces use O_NONBLOCK flag (Greg, Roman & Tejun)

 Documentation/admin-guide/cgroup-v2.rst | 14 ++++++++++++++
 mm/memcontrol.c                         | 10 ++++++++--
 2 files changed, 22 insertions(+), 2 deletions(-)

Setting the max and high limits can trigger synchronous reclaim and/or oom-kill if the usage is higher than the given limit. This behavior is fine for newly created cgroups but it can cause issues for the node controller while setting limits for existing cgroups.

In our production multi-tenant and overcommitted environment, we are seeing priority inversion when the node controller dynamically adjusts the limits of running jobs of different priorities. Based on the system situation, the node controller may reduce the limits of lower priority jobs and increase the limits of higher priority jobs. However we are seeing node controller getting stuck for long period of time while reclaiming from lower priority jobs while setting their limits and also spends a lot of its own CPU.

One of the workaround we are trying is to fork a new process which sets the limit of the lower priority job along with setting an alarm to get itself killed if it get stuck in the reclaim for lower priority job. However we are finding it very unreliable and costly. Either we need a good enough time buffer for the alarm to be delivered after setting limit and potentialy spend a lot of CPU in the reclaim or be unreliable in setting the limit for much shorter but cheaper (less reclaim) alarms.

Let's introduce new limit setting option which does not trigger reclaim and/or oom-kill and let the processes in the target cgroup to trigger reclaim and/or throttling and/or oom-kill in their next charge request. This will make the node controller on multi-tenant overcommitted environment much more reliable.

Signed-off-by Shakeel Butt <[email protected]>

---
Changes since v1:
- Instead of new interfaces use O_NONBLOCK flag (Greg, Roman & Tejun)

 Documentation/admin-guide/cgroup-v2.rst | 14 ++++++++++++++
 mm/memcontrol.c                         | 10 ++++++++--
 2 files changed, 22 insertions(+), 2 deletions(-)

Documentation/admin-guide/cgroup-v2.rst +0 -0

index 8fb14ffab7d1..c14514da4d9a 100644

Documentation/admin-guide/cgroup-v2.rst +12 -0

+        need to dynamically adjust the job's memory limits without
+        expending their own CPU resources on memory reclamation. The
+        job will trigger the reclaim and/or get throttled on its
+        next charge request.
+
   memory.max
 	A read-write single value file which exists on non-root
 	cgroups.  The default is "max".

@@ -1316,6 +1323,13 @@ PAGE_SIZE multiple when read back.
 	Caller could retry them differently, return into userspace
 	as -ENOMEM or silently ignore in cases like disk readahead.
 
+        If memory.max is opened with O_NONBLOCK, then the synchronous
+        reclaim and oom-kill are bypassed. This is useful for admin
+        processes that need to dynamically adjust the job's memory limits
+        without expending their own CPU resources on memory reclamation.
+        The job will trigger the reclaim and/or oom-kill on its next
+        charge request.
+
   memory.reclaim
 	A write-only nested-keyed file which exists for all cgroups.
 

mm/memcontrol.c +8 -2

index 5e2ea8b8a898..6f7362a7756a 100644
@@ -4252,6 +4252,9 @@ static ssize_t memory_high_write(struct kernfs_open_file *of,
 
 	page_counter_set_high(&memcg->memory, high);
 
+	if (of->file->f_flags & O_NONBLOCK)
+		goto out;
+
 	for (;;) {
 		unsigned long nr_pages = page_counter_read(&memcg->memory);
 		unsigned long reclaimed;

@@ -4274,7 +4277,7 @@ static ssize_t memory_high_write(struct kernfs_open_file *of,
 		if (!reclaimed && !nr_retries--)
 			break;
 	}
-
+out:
 	memcg_wb_domain_size_changed(memcg);
 	return nbytes;
 }

@@ -4301,6 +4304,9 @@ static ssize_t memory_max_write(struct kernfs_open_file *of,
 
 	xchg(&memcg->memory.max, max);
 
+	if (of->file->f_flags & O_NONBLOCK)
+		goto out;
+
 	for (;;) {
 		unsigned long nr_pages = page_counter_read(&memcg->memory);
 

@@ -4328,7 +4334,7 @@ static ssize_t memory_max_write(struct kernfs_open_file *of,
 			break;
 		cond_resched();
 	}
-
+out:
 	memcg_wb_domain_size_changed(memcg);
 	return nbytes;
 }

Signature · 1 line

2.47.1

As written this isn't restricted to admin processes though, no? So any unprivileged container can open that file O_NONBLOCK and avoid synchronous reclaim?

Which might be fine I have no idea but it's something to explicitly point out

Quoted 4 lines · Christian Brauner

As written this isn't restricted to admin processes though, no? So any unprivileged container can open that file O_NONBLOCK and avoid synchronous reclaim?

Quoted 2 lines

Which might be fine I have no idea but it's something to explicitly point out

It occurred to me as well but I think this is fine -- changing the limits of a container is (should be) a privileged operation already (ensured by file permissions at opening).
IOW, this doesn't allow bypassing the limits to anyone who couldn't have been able to change them already.

Quoted 5 lines · Michal Koutný

Quoted 4 lines · Christian Brauner

As written this isn't restricted to admin processes though, no? So any unprivileged container can open that file O_NONBLOCK and avoid synchronous reclaim?

Quoted 8 lines

Quoted 2 lines

Which might be fine I have no idea but it's something to explicitly point out

It occurred to me as well but I think this is fine -- changing the limits of a container is (should be) a privileged operation already (ensured by file permissions at opening).
IOW, this doesn't allow bypassing the limits to anyone who couldn't have been able to change them already.

Hm, can you explain what you mean by a privileged operation here? If I have nested containers with user namespaces with delegated cgroup tress, i.e., chowned to them and then some PID 1 or privileged container _within the user namespace_ lowers the limit and uses O_NONBLOCK then it won't trigger synchronous reclaim. Again, this might all be fine I'm just trying to understand.

Setting the max and high limits can trigger synchronous reclaim and/or oom-kill if the usage is higher than the given limit. This behavior is fine for newly created cgroups but it can cause issues for the node controller while setting limits for existing cgroups.

In our production multi-tenant and overcommitted environment, we are seeing priority inversion when the node controller dynamically adjusts the limits of running jobs of different priorities. Based on the system situation, the node controller may reduce the limits of lower priority jobs and increase the limits of higher priority jobs. However we are seeing node controller getting stuck for long period of time while reclaiming from lower priority jobs while setting their limits and also spends a lot of its own CPU.

One of the workaround we are trying is to fork a new process which sets the limit of the lower priority job along with setting an alarm to get itself killed if it get stuck in the reclaim for lower priority job. However we are finding it very unreliable and costly. Either we need a good enough time buffer for the alarm to be delivered after setting limit and potentialy spend a lot of CPU in the reclaim or be unreliable in setting the limit for much shorter but cheaper (less reclaim) alarms.

Let's introduce new limit setting option which does not trigger reclaim and/or oom-kill and let the processes in the target cgroup to trigger reclaim and/or throttling and/or oom-kill in their next charge request. This will make the node controller on multi-tenant overcommitted environment much more reliable.

Signed-off-by Shakeel Butt <[email protected]>