mm: Throttle allocators when failing reclaim over memory.high
Subsystems (3)
PATCH 1/1 mm: Throttle allocators when failing reclaim over memory.high
We're trying to use memory.high to limit workloads, but have found that containment can frequently fail completely and cause OOM situations outside of the cgroup. This happens especially with swap space -- either when none is configured, or swap is full. These failures often also don't have enough warning to allow one to react, whether for a human or for a daemon monitoring PSI.
Here is output from a simple program showing how long it takes in μsec (column 2) to allocate a megabyte of anonymous memory (column 1) when a cgroup is already beyond its memory high setting, and no swap is available:
[root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \
> --wait -t timeout 300 /root/mdf
[...]
95 1035
96 1038
97 1000
98 1036
99 1048
100 1590
101 1968
102 1776
103 1863
104 1757
105 1921
106 1893
107 1760
108 1748
109 1843
110 1716
111 1924
112 1776
113 1831
114 1766
115 1836
116 1588
117 1912
118 1802
119 1857
120 1731
[...]
[System OOM in 2-3 seconds]The delay does go up extremely marginally past the 100MB memory.high threshold, as now we spend time scanning before returning to usermode, but it's nowhere near enough to contain growth. It also doesn't get worse the more pages you have, since it only considers nr_pages.
The current situation goes against both the expectations of users of memory.high, and our intentions as cgroup v2 developers. In cgroup-v2.txt, we claim that we will throttle and only under "extreme conditions" will memory.high protection be breached. Likewise, cgroup v2 users generally also expect that memory.high should throttle workloads as they exceed their high threshold. However, as seen above, this isn't always how it works in practice -- even on banal setups like those with no swap, or where swap has become exhausted, we can end up with memory.high being breached and us having no weapons left in our arsenal to combat runaway growth with, since reclaim is futile.
It's also hard for system monitoring software or users to tell how bad the situation is, as "high" events for the memcg may in some cases be benign, and in others be catastrophic. The current status quo is that we fail containment in a way that doesn't provide any advance warning that things are about to go horribly wrong (for example, we are about to invoke the kernel OOM killer).
This patch introduces explicit throttling when reclaim is failing to keep memcg size contained at the memory.high setting. It does so by applying an exponential delay curve derived from the memcg's overage compared to memory.high. In the normal case where the memcg is either below or only marginally over its memory.high setting, no throttling will be performed.
This composes well with system health monitoring and remediation, as these allocator delays are factored into PSI's memory pressure calculations. This both creates a mechanism system administrators or applications consuming the PSI interface to trivially see that the memcg in question is struggling and use that to make more reasonable decisions, and permits them enough time to act. Either of these can act with significantly more nuance than that we can provide using the system OOM killer.
This is a similar idea to memory.oom_control in cgroup v1 which would put the cgroup to sleep if the threshold was violated, but it's also significantly improved as it results in visible memory pressure, and also doesn't schedule indefinitely, which previously made tracing and other introspection difficult.
Contrast the previous results with a kernel with this patch:
[root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \
> --wait -t timeout 300 /root/mdf
[...]
95 1002
96 1000
97 1002
98 1003
99 1000
100 1043
101 84724
102 330628
103 610511
104 1016265
105 1503969
106 2391692
107 2872061
108 3248003
109 4791904
110 5759832
111 6912509
112 8127818
113 9472203
114 12287622
115 12480079
116 14144008
117 15808029
118 16384500
119 16383242
120 16384979
[...]As you can see, in the normal case, memory allocation takes around 1000 μsec. However, as we exceed our memory.high, things start to increase exponentially, but fairly leniently at first. Our first megabyte over memory.high takes us 0.16 seconds, then the next is 0.46 seconds, then the next is almost an entire second. This gets worse until we reach our eventual 2*HZ clamp per batch, resulting in 16 seconds per megabyte. However, this is still making forward progress, so permits tracing or further analysis with programs like GDB.
This patch expands on earlier work by Johannes Weiner. Thanks!
--- mm/memcontrol.c | 118 +++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 117 insertions(+), 1 deletion(-)
mm/memcontrol.c +117 -1
@@ -65,6 +65,7 @@#include <linux/lockdep.h> #include <linux/file.h> #include <linux/tracehook.h> +#include <linux/psi.h> #include "internal.h" #include <net/sock.h> #include <net/ip.h>
@@ -2161,12 +2162,68 @@ static void high_work_func(struct work_struct *work)reclaim_high (memcg,MEMCG_CHARGE_BATCH ,GFP_KERNEL ); } +/* +* Clamp the maximum sleep time per allocation batch to 2 seconds. This is +* enough to still cause a significant slowdown in most cases, while still +* allowing diagnostics and tracing to proceed without becoming stuck. +*/ +#define MEMCG_MAX_HIGH_DELAY_JIFFIES (2UL*HZ) + +/* +* When calculating the delay, we use these either side of the exponentiation to +* maintain precision and scale to a reasonable number of jiffies (see the table +* below. +* +* - MEMCG_DELAY_PRECISION_SHIFT: Extra precision bits while translating the +* overage ratio to a delay. +* - MEMCG_DELAY_SCALING_SHIFT: The number of bits to scale down down the +* proposed penalty in order to reduce to a reasonable number of jiffies, and +* to produce a reasonable delay curve. +* +* MEMCG_DELAY_SCALING_SHIFT just happens to be a number that produces a +* reasonable delay curve compared to precision-adjusted overage, not +* penalising heavily at first, but still making sure that growth beyond the +* limit penalises misbehaviour cgroups by slowing them down exponentially. For +* example, with a high of 100 megabytes: +* +* +-------+------------------------+ +* | usage | time to allocate in ms | +* +-------+------------------------+ +* | 100M | 0 | +* | 101M | 6 | +* | 102M | 25 | +* | 103M | 57 | +* | 104M | 102 | +* | 105M | 159 | +* | 106M | 230 | +* | 107M | 313 | +* | 108M | 409 | +* | 109M | 518 | +* | 110M | 639 | +* | 111M | 774 | +* | 112M | 921 | +* | 113M | 1081 | +* | 114M | 1254 | +* | 115M | 1439 | +* | 116M | 1638 | +* | 117M | 1849 | +* | 118M | 2000 | +* | 119M | 2000 | +* | 120M | 2000 | +* +-------+------------------------+ +*/ +#define MEMCG_DELAY_PRECISION_SHIFT 20 +#define MEMCG_DELAY_SCALING_SHIFT 14 +/* * Scheduled by try_charge() to be executed from the userland return path * and reclaims memory over the high limit. */ void mem_cgroup_handle_over_high (void ) { +unsigned long usage, high; +unsigned long pflags; +unsigned long penalty_jiffies, overage;unsigned int nr_pages = current->memcg_nr_pages_over_high;struct mem_cgroup *memcg = current->memcg_high_reclaim;
@@ -2177,9 +2234,68 @@ void mem_cgroup_handle_over_high(void) memcg =get_mem_cgroup_from_mm (current->mm);reclaim_high (memcg, nr_pages,GFP_KERNEL ); -css_put (&memcg->css); current->memcg_high_reclaim =NULL ; current->memcg_nr_pages_over_high =0 ; + +/* +* memory.high is breached and reclaim is unable to keep up. Throttle +* allocators proactively to slow down excessive growth. +* +* We use overage compared to memory.high to calculate the number of +* jiffies to sleep (penalty_jiffies). Ideally this value should be +* fairly lenient on small overages, and increasingly harsh when the +* memcg in question makes it clear that it has no intention of stopping +* its crazy behaviour, so we exponentially increase the delay based on +* overage amount. +*/ + + usage =page_counter_read (&memcg->memory); + high =READ_ONCE (memcg->high); + +if (usage <= high) + goto out; + + overage = ((u64)(usage - high) <<MEMCG_DELAY_PRECISION_SHIFT ) / high; + penalty_jiffies = ((u64 )overage * overage *HZ ) + >> (MEMCG_DELAY_PRECISION_SHIFT +MEMCG_DELAY_SCALING_SHIFT ); + +/* +* Factor in the task's own contribution to the overage, such that four +* N-sized allocations are throttled approximately the same as one +* 4N-sized allocation. +* +* MEMCG_CHARGE_BATCH pages is nominal, so work out how much smaller or +* larger the current charge patch is than that. +*/ + penalty_jiffies = penalty_jiffies * nr_pages /MEMCG_CHARGE_BATCH ; + +/* +* Clamp the max delay per usermode return so as to still keep the +* application moving forwards and also permit diagnostics, albeit +* extremely slowly. +*/ + penalty_jiffies =min (penalty_jiffies,MEMCG_MAX_HIGH_DELAY_JIFFIES ); + +/* +* Don't sleep if the amount of jiffies this memcg owes us is so low +* that it's not even worth doing, in an attempt to be nice to those who +* go only a small amount over their memory.high value and maybe haven't +* been aggressively reclaimed enough yet. +*/ +if (penalty_jiffies <=HZ /100 ) + goto out; + +/* +* If we exit early, we're guaranteed to die (since +* schedule_timeout_killable sets TASK_KILLABLE). This means we don't +* need to account for any ill-begotten jiffies to pay them off later. +*/ +psi_memstall_enter (&pflags); +schedule_timeout_killable (penalty_jiffies); +psi_memstall_leave (&pflags); + +out: +css_put (&memcg->css); } static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
Signature · 1 line
2.20.1
Michal Hocko
maintainerOn Thu 31-01-19 20:13:52, Chris Down wrote:
[...]
Quoted 18 lines
The current situation goes against both the expectations of users of memory.high, and our intentions as cgroup v2 developers. In cgroup-v2.txt, we claim that we will throttle and only under "extreme conditions" will memory.high protection be breached. Likewise, cgroup v2 users generally also expect that memory.high should throttle workloads as they exceed their high threshold. However, as seen above, this isn't always how it works in practice -- even on banal setups like those with no swap, or where swap has become exhausted, we can end up with memory.high being breached and us having no weapons left in our arsenal to combat runaway growth with, since reclaim is futile.
It's also hard for system monitoring software or users to tell how bad the situation is, as "high" events for the memcg may in some cases be benign, and in others be catastrophic. The current status quo is that we fail containment in a way that doesn't provide any advance warning that things are about to go horribly wrong (for example, we are about to invoke the kernel OOM killer).
Quoted 6 lines
This patch introduces explicit throttling when reclaim is failing to keep memcg size contained at the memory.high setting. It does so by applying an exponential delay curve derived from the memcg's overage compared to memory.high. In the normal case where the memcg is either below or only marginally over its memory.high setting, no throttling will be performed.
How does this play wit the actual OOM when the user expects oom to resolve the situation because the reclaim is futile and there is nothing reclaimable except for killing a process?
Signature · 2 lines
Michal Hocko SUSE Labs
Johannes Weiner
maintainerQuoted 27 lines · Michal Hocko
On Thu 31-01-19 20:13:52, Chris Down wrote:
[...]Quoted 18 lines
The current situation goes against both the expectations of users of memory.high, and our intentions as cgroup v2 developers. In cgroup-v2.txt, we claim that we will throttle and only under "extreme conditions" will memory.high protection be breached. Likewise, cgroup v2 users generally also expect that memory.high should throttle workloads as they exceed their high threshold. However, as seen above, this isn't always how it works in practice -- even on banal setups like those with no swap, or where swap has become exhausted, we can end up with memory.high being breached and us having no weapons left in our arsenal to combat runaway growth with, since reclaim is futile.
It's also hard for system monitoring software or users to tell how bad the situation is, as "high" events for the memcg may in some cases be benign, and in others be catastrophic. The current status quo is that we fail containment in a way that doesn't provide any advance warning that things are about to go horribly wrong (for example, we are about to invoke the kernel OOM killer).
Quoted 6 lines
This patch introduces explicit throttling when reclaim is failing to keep memcg size contained at the memory.high setting. It does so by applying an exponential delay curve derived from the memcg's overage compared to memory.high. In the normal case where the memcg is either below or only marginally over its memory.high setting, no throttling will be performed.
Quoted 3 lines
How does this play wit the actual OOM when the user expects oom to resolve the situation because the reclaim is futile and there is nothing reclaimable except for killing a process?
Hm, can you elaborate on your question a bit?
The idea behind memory.high is to throttle allocations long enough for the admin or a management daemon to intervene, but not to trigger the kernel oom killer. It was designed as a replacement for the cgroup1 oom_control, but without the deadlock potential, ptrace problems etc.
What we specifically do is to set memory.high and have a daemon (oomd) watch memory.pressure, io.pressure etc. in the group. If pressure exceeds a certain threshold, the daemon kills something.
As you know, the kernel OOM killer does not kick in reliably when e.g. page cache is thrashing heavily, since from a kernel POV it's still successfully allocating and reclaiming - meanwhile the workload is spending most its time in page faults. And when the kernel OOM killer does kick in, its selection policy is not very workload-aware.
This daemon on the other hand can be configured to 1) kick in reliably when the workload-specific tolerances for slowdowns and latencies are violated (which tends to be way earlier than the kernel oom killer usually kicks in) and 2) know about the workload and all its components to make an informed kill decision.
Right now, that throttling mechanism works okay with swap enabled, but we cannot enable swap everywhere, or sometimes run out of swap, and then it breaks down and we run into system OOMs.
This patch makes sure memory.high *always* implements the throttling semantics described in cgroup-v2.txt, not just most of the time.
Quoted 3 lines · Michal Hocko
How does this play wit the actual OOM when the user expects oom to resolve the situation because the reclaim is futile and there is nothing reclaimable except for killing a process?
In addition to what Johannes said, this doesn't impede OOM in the case of global system starvation (eg. in the case that all major consumers of memory are allocator throttling). In that case nothing unusual will happen, since the task's state is TASK_KILLABLE rather than TASK_UNINTERRUPTIBLE, and we will exit out of mem_cgroup_handle_over_high as quickly as possible.
Michal Hocko
maintainer[Sorry for a late reply]
Quoted 32 lines · Johannes Weiner
Quoted 27 lines · Michal Hocko
On Thu 31-01-19 20:13:52, Chris Down wrote:
[...]Quoted 18 lines
The current situation goes against both the expectations of users of memory.high, and our intentions as cgroup v2 developers. In cgroup-v2.txt, we claim that we will throttle and only under "extreme conditions" will memory.high protection be breached. Likewise, cgroup v2 users generally also expect that memory.high should throttle workloads as they exceed their high threshold. However, as seen above, this isn't always how it works in practice -- even on banal setups like those with no swap, or where swap has become exhausted, we can end up with memory.high being breached and us having no weapons left in our arsenal to combat runaway growth with, since reclaim is futile.
It's also hard for system monitoring software or users to tell how bad the situation is, as "high" events for the memcg may in some cases be benign, and in others be catastrophic. The current status quo is that we fail containment in a way that doesn't provide any advance warning that things are about to go horribly wrong (for example, we are about to invoke the kernel OOM killer).
Quoted 6 lines
This patch introduces explicit throttling when reclaim is failing to keep memcg size contained at the memory.high setting. It does so by applying an exponential delay curve derived from the memcg's overage compared to memory.high. In the normal case where the memcg is either below or only marginally over its memory.high setting, no throttling will be performed.
Quoted 3 lines
How does this play wit the actual OOM when the user expects oom to resolve the situation because the reclaim is futile and there is nothing reclaimable except for killing a process?
Quoted 6 lines
Hm, can you elaborate on your question a bit?
The idea behind memory.high is to throttle allocations long enough for the admin or a management daemon to intervene, but not to trigger the kernel oom killer. It was designed as a replacement for the cgroup1
oom_control, but without the deadlock potential, ptrace problems etc.
Yes, this makes sense. The high limit reclaim is also documented as a best effort resource guarantee. My understanding is that if the workload cannot be contained within the high limit then the system cannot do much and eventually gives up. Having the full memory unreclaimable is such an example. And there is either the global OOM killer or hard limit OOM killer to trigger to resolve such a situation.
[...]
Thanks for describing the usecase.
Quoted 6 lines
Right now, that throttling mechanism works okay with swap enabled, but we cannot enable swap everywhere, or sometimes run out of swap, and then it breaks down and we run into system OOMs.
This patch makes sure memory.high *always* implements the throttling semantics described in cgroup-v2.txt, not just most of the time.
I am not really opposed to the throttling in the absence of a reclaimable memory. We do that for the regular allocation paths already (should_reclaim_retry). A swapless system with anon memory is very likely to oom too quickly and this sounds like a real problem. But I do not think that we should throttle the allocation to freeze it completely. We should eventually OOM. And that was my question about essentially. How much we can/should throttle to give a high limit events consumer enough time to intervene. I am sorry to still not have time to study the patch more closely but this should be explained in the changelog. Are we talking about seconds/minutes or simply freeze each allocator to death?
Signature · 2 lines
Michal Hocko SUSE Labs
PATCH 1/1 mm: Throttle allocators when failing reclaim over memory.high
In reply to Chris DownWe're trying to use memory.high to limit workloads, but have found that containment can frequently fail completely and cause OOM situations outside of the cgroup. This happens especially with swap space -- either when none is configured, or swap is full. These failures often also don't have enough warning to allow one to react, whether for a human or for a daemon monitoring PSI.
Here is output from a simple program showing how long it takes in μsec (column 2) to allocate a megabyte of anonymous memory (column 1) when a cgroup is already beyond its memory high setting, and no swap is available:
[root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \
> --wait -t timeout 300 /root/mdf
[...]
95 1035
96 1038
97 1000
98 1036
99 1048
100 1590
101 1968
102 1776
103 1863
104 1757
105 1921
106 1893
107 1760
108 1748
109 1843
110 1716
111 1924
112 1776
113 1831
114 1766
115 1836
116 1588
117 1912
118 1802
119 1857
120 1731
[...]
[System OOM in 2-3 seconds]The delay does go up extremely marginally past the 100MB memory.high threshold, as now we spend time scanning before returning to usermode, but it's nowhere near enough to contain growth. It also doesn't get worse the more pages you have, since it only considers nr_pages.
The current situation goes against both the expectations of users of memory.high, and our intentions as cgroup v2 developers. In cgroup-v2.txt, we claim that we will throttle and only under "extreme conditions" will memory.high protection be breached. Likewise, cgroup v2 users generally also expect that memory.high should throttle workloads as they exceed their high threshold. However, as seen above, this isn't always how it works in practice -- even on banal setups like those with no swap, or where swap has become exhausted, we can end up with memory.high being breached and us having no weapons left in our arsenal to combat runaway growth with, since reclaim is futile.
It's also hard for system monitoring software or users to tell how bad the situation is, as "high" events for the memcg may in some cases be benign, and in others be catastrophic. The current status quo is that we fail containment in a way that doesn't provide any advance warning that things are about to go horribly wrong (for example, we are about to invoke the kernel OOM killer).
This patch introduces explicit throttling when reclaim is failing to keep memcg size contained at the memory.high setting. It does so by applying an exponential delay curve derived from the memcg's overage compared to memory.high. In the normal case where the memcg is either below or only marginally over its memory.high setting, no throttling will be performed.
This composes well with system health monitoring and remediation, as these allocator delays are factored into PSI's memory pressure calculations. This both creates a mechanism system administrators or applications consuming the PSI interface to trivially see that the memcg in question is struggling and use that to make more reasonable decisions, and permits them enough time to act. Either of these can act with significantly more nuance than that we can provide using the system OOM killer.
This is a similar idea to memory.oom_control in cgroup v1 which would put the cgroup to sleep if the threshold was violated, but it's also significantly improved as it results in visible memory pressure, and also doesn't schedule indefinitely, which previously made tracing and other introspection difficult.
Contrast the previous results with a kernel with this patch:
[root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \
> --wait -t timeout 300 /root/mdf
[...]
95 1002
96 1000
97 1002
98 1003
99 1000
100 1043
101 84724
102 330628
103 610511
104 1016265
105 1503969
106 2391692
107 2872061
108 3248003
109 4791904
110 5759832
111 6912509
112 8127818
113 9472203
114 12287622
115 12480079
116 14144008
117 15808029
118 16384500
119 16383242
120 16384979
[...]As you can see, in the normal case, memory allocation takes around 1000 μsec. However, as we exceed our memory.high, things start to increase exponentially, but fairly leniently at first. Our first megabyte over memory.high takes us 0.16 seconds, then the next is 0.46 seconds, then the next is almost an entire second. This gets worse until we reach our eventual 2*HZ clamp per batch, resulting in 16 seconds per megabyte. However, this is still making forward progress, so permits tracing or further analysis with programs like GDB.
This patch expands on earlier work by Johannes Weiner. Thanks!
--- mm/memcontrol.c | 118 +++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 117 insertions(+), 1 deletion(-)
mm/memcontrol.c +117 -1
@@ -66,6 +66,7 @@#include <linux/lockdep.h> #include <linux/file.h> #include <linux/tracehook.h> +#include <linux/psi.h> #include "internal.h" #include <net/sock.h> #include <net/ip.h>
@@ -2142,12 +2143,68 @@ static void high_work_func(struct work_struct *work)reclaim_high (memcg,MEMCG_CHARGE_BATCH ,GFP_KERNEL ); } +/* +* Clamp the maximum sleep time per allocation batch to 2 seconds. This is +* enough to still cause a significant slowdown in most cases, while still +* allowing diagnostics and tracing to proceed without becoming stuck. +*/ +#define MEMCG_MAX_HIGH_DELAY_JIFFIES (2UL*HZ) + +/* +* When calculating the delay, we use these either side of the exponentiation to +* maintain precision and scale to a reasonable number of jiffies (see the table +* below. +* +* - MEMCG_DELAY_PRECISION_SHIFT: Extra precision bits while translating the +* overage ratio to a delay. +* - MEMCG_DELAY_SCALING_SHIFT: The number of bits to scale down down the +* proposed penalty in order to reduce to a reasonable number of jiffies, and +* to produce a reasonable delay curve. +* +* MEMCG_DELAY_SCALING_SHIFT just happens to be a number that produces a +* reasonable delay curve compared to precision-adjusted overage, not +* penalising heavily at first, but still making sure that growth beyond the +* limit penalises misbehaviour cgroups by slowing them down exponentially. For +* example, with a high of 100 megabytes: +* +* +-------+------------------------+ +* | usage | time to allocate in ms | +* +-------+------------------------+ +* | 100M | 0 | +* | 101M | 6 | +* | 102M | 25 | +* | 103M | 57 | +* | 104M | 102 | +* | 105M | 159 | +* | 106M | 230 | +* | 107M | 313 | +* | 108M | 409 | +* | 109M | 518 | +* | 110M | 639 | +* | 111M | 774 | +* | 112M | 921 | +* | 113M | 1081 | +* | 114M | 1254 | +* | 115M | 1439 | +* | 116M | 1638 | +* | 117M | 1849 | +* | 118M | 2000 | +* | 119M | 2000 | +* | 120M | 2000 | +* +-------+------------------------+ +*/ +#define MEMCG_DELAY_PRECISION_SHIFT 20 +#define MEMCG_DELAY_SCALING_SHIFT 14 +/* * Scheduled by try_charge() to be executed from the userland return path * and reclaims memory over the high limit. */ void mem_cgroup_handle_over_high (void ) { +unsigned long usage, high; +unsigned long pflags; +unsigned long penalty_jiffies, overage;unsigned int nr_pages = current->memcg_nr_pages_over_high;struct mem_cgroup *memcg = current->memcg_high_reclaim;
@@ -2158,9 +2215,68 @@ void mem_cgroup_handle_over_high(void) memcg =get_mem_cgroup_from_mm (current->mm);reclaim_high (memcg, nr_pages,GFP_KERNEL ); -css_put (&memcg->css); current->memcg_high_reclaim =NULL ; current->memcg_nr_pages_over_high =0 ; + +/* +* memory.high is breached and reclaim is unable to keep up. Throttle +* allocators proactively to slow down excessive growth. +* +* We use overage compared to memory.high to calculate the number of +* jiffies to sleep (penalty_jiffies). Ideally this value should be +* fairly lenient on small overages, and increasingly harsh when the +* memcg in question makes it clear that it has no intention of stopping +* its crazy behaviour, so we exponentially increase the delay based on +* overage amount. +*/ + + usage =page_counter_read (&memcg->memory); + high =READ_ONCE (memcg->high); + +if (usage <= high) + goto out; + + overage = ((u64)(usage - high) <<MEMCG_DELAY_PRECISION_SHIFT ) / high; + penalty_jiffies = ((u64 )overage * overage *HZ ) + >> (MEMCG_DELAY_PRECISION_SHIFT +MEMCG_DELAY_SCALING_SHIFT ); + +/* +* Factor in the task's own contribution to the overage, such that four +* N-sized allocations are throttled approximately the same as one +* 4N-sized allocation. +* +* MEMCG_CHARGE_BATCH pages is nominal, so work out how much smaller or +* larger the current charge patch is than that. +*/ + penalty_jiffies = penalty_jiffies * nr_pages /MEMCG_CHARGE_BATCH ; + +/* +* Clamp the max delay per usermode return so as to still keep the +* application moving forwards and also permit diagnostics, albeit +* extremely slowly. +*/ + penalty_jiffies =min (penalty_jiffies,MEMCG_MAX_HIGH_DELAY_JIFFIES ); + +/* +* Don't sleep if the amount of jiffies this memcg owes us is so low +* that it's not even worth doing, in an attempt to be nice to those who +* go only a small amount over their memory.high value and maybe haven't +* been aggressively reclaimed enough yet. +*/ +if (penalty_jiffies <=HZ /100 ) + goto out; + +/* +* If we exit early, we're guaranteed to die (since +* schedule_timeout_killable sets TASK_KILLABLE). This means we don't +* need to account for any ill-begotten jiffies to pay them off later. +*/ +psi_memstall_enter (&pflags); +schedule_timeout_killable (penalty_jiffies); +psi_memstall_leave (&pflags); + +out: +css_put (&memcg->css); } static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
Signature · 1 line
2.21.0
Hey Michal,
Just to come back to your last e-mail about how this interacts with OOM.
Quoted 10 lines · Michal Hocko
I am not really opposed to the throttling in the absence of a reclaimable memory. We do that for the regular allocation paths already (
should_reclaim_retry). A swapless system with anon memory is very likely to oom too quickly and this sounds like a real problem. But I do not think that we should throttle the allocation to freeze it completely. We should eventually OOM. And that was my question about essentially. How much we can/should throttle to give a high limit events consumer enough time to intervene. I am sorry to still not have time to study the patch more closely but this should be explained in the changelog. Are we talking about seconds/minutes or simply freeze each allocator to death?
Per-allocation, the maximum is 2 seconds (MEMCG_MAX_HIGH_DELAY_JIFFIES), so we don't freeze things to death -- they can recover if they are amenable to it. The idea here is that primarily you handle it, just like memory.oom_control in v1 (as mentioned in the commit message, or as a last resort, the kernel will still OOM if our userspace daemon has kicked the bucket or is otherwise ineffective.
If you're setting memory.high and memory.max together, then setting memory.high always has to come with a.) tolerance of heavy throttling by your application, and b.) userspace intervention in the case of high memory pressure resulting. This patch doesn't really change those semantics.
PATCH v3 1/1 mm: Throttle allocators when failing reclaim over memory.high
We're trying to use memory.high to limit workloads, but have found that containment can frequently fail completely and cause OOM situations outside of the cgroup. This happens especially with swap space -- either when none is configured, or swap is full. These failures often also don't have enough warning to allow one to react, whether for a human or for a daemon monitoring PSI.
Here is output from a simple program showing how long it takes in μsec (column 2) to allocate a megabyte of anonymous memory (column 1) when a cgroup is already beyond its memory high setting, and no swap is available:
[root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \
> --wait -t timeout 300 /root/mdf
[...]
95 1035
96 1038
97 1000
98 1036
99 1048
100 1590
101 1968
102 1776
103 1863
104 1757
105 1921
106 1893
107 1760
108 1748
109 1843
110 1716
111 1924
112 1776
113 1831
114 1766
115 1836
116 1588
117 1912
118 1802
119 1857
120 1731
[...]
[System OOM in 2-3 seconds]The delay does go up extremely marginally past the 100MB memory.high threshold, as now we spend time scanning before returning to usermode, but it's nowhere near enough to contain growth. It also doesn't get worse the more pages you have, since it only considers nr_pages.
The current situation goes against both the expectations of users of memory.high, and our intentions as cgroup v2 developers. In cgroup-v2.txt, we claim that we will throttle and only under "extreme conditions" will memory.high protection be breached. Likewise, cgroup v2 users generally also expect that memory.high should throttle workloads as they exceed their high threshold. However, as seen above, this isn't always how it works in practice -- even on banal setups like those with no swap, or where swap has become exhausted, we can end up with memory.high being breached and us having no weapons left in our arsenal to combat runaway growth with, since reclaim is futile.
It's also hard for system monitoring software or users to tell how bad the situation is, as "high" events for the memcg may in some cases be benign, and in others be catastrophic. The current status quo is that we fail containment in a way that doesn't provide any advance warning that things are about to go horribly wrong (for example, we are about to invoke the kernel OOM killer).
This patch introduces explicit throttling when reclaim is failing to keep memcg size contained at the memory.high setting. It does so by applying an exponential delay curve derived from the memcg's overage compared to memory.high. In the normal case where the memcg is either below or only marginally over its memory.high setting, no throttling will be performed.
This composes well with system health monitoring and remediation, as these allocator delays are factored into PSI's memory pressure calculations. This both creates a mechanism system administrators or applications consuming the PSI interface to trivially see that the memcg in question is struggling and use that to make more reasonable decisions, and permits them enough time to act. Either of these can act with significantly more nuance than that we can provide using the system OOM killer.
This is a similar idea to memory.oom_control in cgroup v1 which would put the cgroup to sleep if the threshold was violated, but it's also significantly improved as it results in visible memory pressure, and also doesn't schedule indefinitely, which previously made tracing and other introspection difficult (ie. it's clamped at 2*HZ per allocation through MEMCG_MAX_HIGH_DELAY_JIFFIES).
Contrast the previous results with a kernel with this patch:
[root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \
> --wait -t timeout 300 /root/mdf
[...]
95 1002
96 1000
97 1002
98 1003
99 1000
100 1043
101 84724
102 330628
103 610511
104 1016265
105 1503969
106 2391692
107 2872061
108 3248003
109 4791904
110 5759832
111 6912509
112 8127818
113 9472203
114 12287622
115 12480079
116 14144008
117 15808029
118 16384500
119 16383242
120 16384979
[...]As you can see, in the normal case, memory allocation takes around 1000 μsec. However, as we exceed our memory.high, things start to increase exponentially, but fairly leniently at first. Our first megabyte over memory.high takes us 0.16 seconds, then the next is 0.46 seconds, then the next is almost an entire second. This gets worse until we reach our eventual 2*HZ clamp per batch, resulting in 16 seconds per megabyte. However, this is still making forward progress, so permits tracing or further analysis with programs like GDB.
We use an exponential curve for our delay penalty for a few reasons:
1. We run mem_cgroup_handle_over_high to potentially do reclaim after we've already performed allocations, which means that temporarily going over memory.high by a small amount may be perfectly legitimate, even for compliant workloads. We don't want to unduly penalise such cases.
2. An exponential curve (as opposed to a static or linear delay) allows ramping up memory pressure stats more gradually, which can be useful to work out that you have set memory.high too low, without destroying application performance entirely.
This patch expands on earlier work by Johannes Weiner. Thanks!
--- mm/memcontrol.c | 118 +++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 117 insertions(+), 1 deletion(-)
[v3: updated the changelog post discussion in person with Michal.]
mm/memcontrol.c +117 -1
@@ -66,6 +66,7 @@#include <linux/lockdep.h> #include <linux/file.h> #include <linux/tracehook.h> +#include <linux/psi.h> #include "internal.h" #include <net/sock.h> #include <net/ip.h>
@@ -2263,12 +2264,68 @@ static void high_work_func(struct work_struct *work)reclaim_high (memcg,MEMCG_CHARGE_BATCH ,GFP_KERNEL ); } +/* +* Clamp the maximum sleep time per allocation batch to 2 seconds. This is +* enough to still cause a significant slowdown in most cases, while still +* allowing diagnostics and tracing to proceed without becoming stuck. +*/ +#define MEMCG_MAX_HIGH_DELAY_JIFFIES (2UL*HZ) + +/* +* When calculating the delay, we use these either side of the exponentiation to +* maintain precision and scale to a reasonable number of jiffies (see the table +* below. +* +* - MEMCG_DELAY_PRECISION_SHIFT: Extra precision bits while translating the +* overage ratio to a delay. +* - MEMCG_DELAY_SCALING_SHIFT: The number of bits to scale down down the +* proposed penalty in order to reduce to a reasonable number of jiffies, and +* to produce a reasonable delay curve. +* +* MEMCG_DELAY_SCALING_SHIFT just happens to be a number that produces a +* reasonable delay curve compared to precision-adjusted overage, not +* penalising heavily at first, but still making sure that growth beyond the +* limit penalises misbehaviour cgroups by slowing them down exponentially. For +* example, with a high of 100 megabytes: +* +* +-------+------------------------+ +* | usage | time to allocate in ms | +* +-------+------------------------+ +* | 100M | 0 | +* | 101M | 6 | +* | 102M | 25 | +* | 103M | 57 | +* | 104M | 102 | +* | 105M | 159 | +* | 106M | 230 | +* | 107M | 313 | +* | 108M | 409 | +* | 109M | 518 | +* | 110M | 639 | +* | 111M | 774 | +* | 112M | 921 | +* | 113M | 1081 | +* | 114M | 1254 | +* | 115M | 1439 | +* | 116M | 1638 | +* | 117M | 1849 | +* | 118M | 2000 | +* | 119M | 2000 | +* | 120M | 2000 | +* +-------+------------------------+ +*/ +#define MEMCG_DELAY_PRECISION_SHIFT 20 +#define MEMCG_DELAY_SCALING_SHIFT 14 +/* * Scheduled by try_charge() to be executed from the userland return path * and reclaims memory over the high limit. */ void mem_cgroup_handle_over_high (void ) { +unsigned long usage, high; +unsigned long pflags; +unsigned long penalty_jiffies, overage;unsigned int nr_pages = current->memcg_nr_pages_over_high;struct mem_cgroup *memcg = current->memcg_high_reclaim;
@@ -2279,9 +2336,68 @@ void mem_cgroup_handle_over_high(void) memcg =get_mem_cgroup_from_mm (current->mm);reclaim_high (memcg, nr_pages,GFP_KERNEL ); -css_put (&memcg->css); current->memcg_high_reclaim =NULL ; current->memcg_nr_pages_over_high =0 ; + +/* +* memory.high is breached and reclaim is unable to keep up. Throttle +* allocators proactively to slow down excessive growth. +* +* We use overage compared to memory.high to calculate the number of +* jiffies to sleep (penalty_jiffies). Ideally this value should be +* fairly lenient on small overages, and increasingly harsh when the +* memcg in question makes it clear that it has no intention of stopping +* its crazy behaviour, so we exponentially increase the delay based on +* overage amount. +*/ + + usage =page_counter_read (&memcg->memory); + high =READ_ONCE (memcg->high); + +if (usage <= high) + goto out; + + overage = ((u64)(usage - high) <<MEMCG_DELAY_PRECISION_SHIFT ) / high; + penalty_jiffies = ((u64 )overage * overage *HZ ) + >> (MEMCG_DELAY_PRECISION_SHIFT +MEMCG_DELAY_SCALING_SHIFT ); + +/* +* Factor in the task's own contribution to the overage, such that four +* N-sized allocations are throttled approximately the same as one +* 4N-sized allocation. +* +* MEMCG_CHARGE_BATCH pages is nominal, so work out how much smaller or +* larger the current charge patch is than that. +*/ + penalty_jiffies = penalty_jiffies * nr_pages /MEMCG_CHARGE_BATCH ; + +/* +* Clamp the max delay per usermode return so as to still keep the +* application moving forwards and also permit diagnostics, albeit +* extremely slowly. +*/ + penalty_jiffies =min (penalty_jiffies,MEMCG_MAX_HIGH_DELAY_JIFFIES ); + +/* +* Don't sleep if the amount of jiffies this memcg owes us is so low +* that it's not even worth doing, in an attempt to be nice to those who +* go only a small amount over their memory.high value and maybe haven't +* been aggressively reclaimed enough yet. +*/ +if (penalty_jiffies <=HZ /100 ) + goto out; + +/* +* If we exit early, we're guaranteed to die (since +* schedule_timeout_killable sets TASK_KILLABLE). This means we don't +* need to account for any ill-begotten jiffies to pay them off later. +*/ +psi_memstall_enter (&pflags); +schedule_timeout_killable (penalty_jiffies); +psi_memstall_leave (&pflags); + +out: +css_put (&memcg->css); } static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
Signature · 1 line
2.21.0
Michal Hocko
maintainerSorry I was traveling last week and will be off for a large part of this week.
Quoted 49 lines · Chris Down
We're trying to use memory.high to limit workloads, but have found that containment can frequently fail completely and cause OOM situations outside of the cgroup. This happens especially with swap space -- either when none is configured, or swap is full. These failures often also don't have enough warning to allow one to react, whether for a human or for a daemon monitoring PSI.
Here is output from a simple program showing how long it takes in μsec (column 2) to allocate a megabyte of anonymous memory (column 1) when a cgroup is already beyond its memory high setting, and no swap is available:
[root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \ > --wait -t timeout 300 /root/mdf [...] 95 1035 96 1038 97 1000 98 1036 99 1048 100 1590 101 1968 102 1776 103 1863 104 1757 105 1921 106 1893 107 1760 108 1748 109 1843 110 1716 111 1924 112 1776 113 1831 114 1766 115 1836 116 1588 117 1912 118 1802 119 1857 120 1731 [...] [System OOM in 2-3 seconds]The delay does go up extremely marginally past the 100MB memory.high threshold, as now we spend time scanning before returning to usermode, but it's nowhere near enough to contain growth. It also doesn't get worse the more pages you have, since it only considers
nr_pages.
Quoted 10 lines
The current situation goes against both the expectations of users of memory.high, and our intentions as cgroup v2 developers. In cgroup-v2.txt, we claim that we will throttle and only under "extreme conditions" will memory.high protection be breached. Likewise, cgroup v2 users generally also expect that memory.high should throttle workloads as they exceed their high threshold. However, as seen above, this isn't always how it works in practice -- even on banal setups like those with no swap, or where swap has become exhausted, we can end up with memory.high being breached and us having no weapons left in our arsenal to combat runaway growth with, since reclaim is futile.
Well, having only a non-reclaimable memory essentially means the workload termination to resolve the situation. Be it an in kernel oom killer or any other pro-active measure doing the same. On the other hand I do understand that we shouldn't run into the system oom situation prematurely because the userspace doesn't have enough time to react. In your example above it takes 1M/1ms and that indeed doesn't give much room to
a potential userspace intervention. The global case already tries to throttle on no progress (albeit only for pending writers). I do not see a fundamental reason to not throttle on no progress here as well.
Quoted 23 lines
It's also hard for system monitoring software or users to tell how bad the situation is, as "high" events for the memcg may in some cases be benign, and in others be catastrophic. The current status quo is that we fail containment in a way that doesn't provide any advance warning that things are about to go horribly wrong (for example, we are about to invoke the kernel OOM killer).
This patch introduces explicit throttling when reclaim is failing to keep memcg size contained at the memory.high setting. It does so by applying an exponential delay curve derived from the memcg's overage compared to memory.high. In the normal case where the memcg is either below or only marginally over its memory.high setting, no throttling will be performed.
This composes well with system health monitoring and remediation, as these allocator delays are factored into PSI's memory pressure calculations. This both creates a mechanism system administrators or applications consuming the PSI interface to trivially see that the memcg in question is struggling and use that to make more reasonable decisions, and permits them enough time to act. Either of these can act with significantly more nuance than that we can provide using the system OOM killer.
Quoted 6 lines
This is a similar idea to memory.oom_control in cgroup v1 which would put the cgroup to sleep if the threshold was violated, but it's also significantly improved as it results in visible memory pressure, and also doesn't schedule indefinitely, which previously made tracing and other introspection difficult (ie. it's clamped at 2*HZ per allocation through
MEMCG_MAX_HIGH_DELAY_JIFFIES).
OK, fair enough. One thing that I really do not want to see is to have this explicitly documented in the user documentation because this is an implementation detail, userspace shouldn't depend on.
Quoted 44 lines
Contrast the previous results with a kernel with this patch:
[root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \ > --wait -t timeout 300 /root/mdf [...] 95 1002 96 1000 97 1002 98 1003 99 1000 100 1043 101 84724 102 330628 103 610511 104 1016265 105 1503969 106 2391692 107 2872061 108 3248003 109 4791904 110 5759832 111 6912509 112 8127818 113 9472203 114 12287622 115 12480079 116 14144008 117 15808029 118 16384500 119 16383242 120 16384979 [...]As you can see, in the normal case, memory allocation takes around 1000 μsec. However, as we exceed our memory.high, things start to increase exponentially, but fairly leniently at first. Our first megabyte over memory.high takes us 0.16 seconds, then the next is 0.46 seconds, then the next is almost an entire second. This gets worse until we reach our eventual 2*HZ clamp per batch, resulting in 16 seconds per megabyte. However, this is still making forward progress, so permits tracing or further analysis with programs like GDB.
We use an exponential curve for our delay penalty for a few reasons:
Quoted 9 lines
1. We run
mem_cgroup_handle_over_highto potentially do reclaim after we've already performed allocations, which means that temporarily going over memory.high by a small amount may be perfectly legitimate, even for compliant workloads. We don't want to unduly penalise such cases.
2. An exponential curve (as opposed to a static or linear delay) allows ramping up memory pressure stats more gradually, which can be useful to work out that you have set memory.high too low, without destroying application performance entirely.
Thanks for extending the changelog!
Quoted 11 lines
This patch expands on earlier work by Johannes Weiner. Thanks!
Signed-off-by Chris Down <[email protected]>Cc Andrew Morton <[email protected]>Cc Johannes Weiner <[email protected]>Cc Tejun Heo <[email protected]>Cc Roman Gushchin <[email protected]>
I didn't get to read the code carefully so I cannot really give my ack but the idea seems plausible and if the current parameters turn out to hurt some workloads we can tune from there. The important part is that throttling doesn't push hard limit triggering to infinity because we need the OOM killer to trigger and resolve the situation eventually as not everybody is doing userspace oom killing.
The other important thing is that this throttling is completely transparent to the userspace and it is a subject of the current implementation rather than any form of contract (unlike the oom_control example or any tunable to control the decay/max timeout etc). No userspace should depend on it.
Thanks
Quoted 159 lines
--- mm/memcontrol.c | 118 +++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 117 insertions(+), 1 deletion(-)[v3: updated the changelog post discussion in person with Michal.]
mm/memcontrol.c +117
-1@@ -66,6 +66,7 @@#include <linux/lockdep.h> #include <linux/file.h> #include <linux/tracehook.h> +#include <linux/psi.h> #include "internal.h" #include <net/sock.h> #include <net/ip.h> @@ -2263,12 +2264,68 @@ static void high_work_func(struct work_struct *work)reclaim_high (memcg,MEMCG_CHARGE_BATCH ,GFP_KERNEL ); } +/* +* Clamp the maximum sleep time per allocation batch to 2 seconds. This is +* enough to still cause a significant slowdown in most cases, while still +* allowing diagnostics and tracing to proceed without becoming stuck. +*/ +#define MEMCG_MAX_HIGH_DELAY_JIFFIES (2UL*HZ) + +/* +* When calculating the delay, we use these either side of the exponentiation to +* maintain precision and scale to a reasonable number of jiffies (see the table +* below. +* +* - MEMCG_DELAY_PRECISION_SHIFT: Extra precision bits while translating the +* overage ratio to a delay. +* - MEMCG_DELAY_SCALING_SHIFT: The number of bits to scale down down the +* proposed penalty in order to reduce to a reasonable number of jiffies, and +* to produce a reasonable delay curve. +* +* MEMCG_DELAY_SCALING_SHIFT just happens to be a number that produces a +* reasonable delay curve compared to precision-adjusted overage, not +* penalising heavily at first, but still making sure that growth beyond the +* limit penalises misbehaviour cgroups by slowing them down exponentially. For +* example, with a high of 100 megabytes: +* +* +-------+------------------------+ +* | usage | time to allocate in ms | +* +-------+------------------------+ +* | 100M | 0 | +* | 101M | 6 | +* | 102M | 25 | +* | 103M | 57 | +* | 104M | 102 | +* | 105M | 159 | +* | 106M | 230 | +* | 107M | 313 | +* | 108M | 409 | +* | 109M | 518 | +* | 110M | 639 | +* | 111M | 774 | +* | 112M | 921 | +* | 113M | 1081 | +* | 114M | 1254 | +* | 115M | 1439 | +* | 116M | 1638 | +* | 117M | 1849 | +* | 118M | 2000 | +* | 119M | 2000 | +* | 120M | 2000 | +* +-------+------------------------+ +*/ +#define MEMCG_DELAY_PRECISION_SHIFT 20 +#define MEMCG_DELAY_SCALING_SHIFT 14 +/* * Scheduled by try_charge() to be executed from the userland return path * and reclaims memory over the high limit. */ void mem_cgroup_handle_over_high (void ) { +unsigned long usage, high; +unsigned long pflags; +unsigned long penalty_jiffies, overage;unsigned int nr_pages = current->memcg_nr_pages_over_high;struct mem_cgroup *memcg = current->memcg_high_reclaim;@@ -2279,9 +2336,68 @@ void mem_cgroup_handle_over_high(void) memcg =get_mem_cgroup_from_mm (current->mm);reclaim_high (memcg, nr_pages,GFP_KERNEL ); -css_put (&memcg->css); current->memcg_high_reclaim =NULL ; current->memcg_nr_pages_over_high =0 ; + +/* +* memory.high is breached and reclaim is unable to keep up. Throttle +* allocators proactively to slow down excessive growth. +* +* We use overage compared to memory.high to calculate the number of +* jiffies to sleep (penalty_jiffies). Ideally this value should be +* fairly lenient on small overages, and increasingly harsh when the +* memcg in question makes it clear that it has no intention of stopping +* its crazy behaviour, so we exponentially increase the delay based on +* overage amount. +*/ + + usage =page_counter_read (&memcg->memory); + high =READ_ONCE (memcg->high); + +if (usage <= high) + goto out; + + overage = ((u64)(usage - high) <<MEMCG_DELAY_PRECISION_SHIFT ) / high; + penalty_jiffies = ((u64 )overage * overage *HZ ) + >> (MEMCG_DELAY_PRECISION_SHIFT +MEMCG_DELAY_SCALING_SHIFT ); + +/* +* Factor in the task's own contribution to the overage, such that four +* N-sized allocations are throttled approximately the same as one +* 4N-sized allocation. +* +* MEMCG_CHARGE_BATCH pages is nominal, so work out how much smaller or +* larger the current charge patch is than that. +*/ + penalty_jiffies = penalty_jiffies * nr_pages /MEMCG_CHARGE_BATCH ; + +/* +* Clamp the max delay per usermode return so as to still keep the +* application moving forwards and also permit diagnostics, albeit +* extremely slowly. +*/ + penalty_jiffies =min (penalty_jiffies,MEMCG_MAX_HIGH_DELAY_JIFFIES ); + +/* +* Don't sleep if the amount of jiffies this memcg owes us is so low +* that it's not even worth doing, in an attempt to be nice to those who +* go only a small amount over their memory.high value and maybe haven't +* been aggressively reclaimed enough yet. +*/ +if (penalty_jiffies <=HZ /100 ) + goto out; + +/* +* If we exit early, we're guaranteed to die (since +* schedule_timeout_killable sets TASK_KILLABLE). This means we don't +* need to account for any ill-begotten jiffies to pay them off later. +*/ +psi_memstall_enter (&pflags); +schedule_timeout_killable (penalty_jiffies); +psi_memstall_leave (&pflags); + +out: +css_put (&memcg->css); } static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,Signature · 1 line
2.21.0
Signature · 2 lines
Michal Hocko SUSE Labs
PATCH v4 1/1 mm: Throttle allocators when failing reclaim over memory.high
In reply to Chris DownWe're trying to use memory.high to limit workloads, but have found that containment can frequently fail completely and cause OOM situations outside of the cgroup. This happens especially with swap space -- either when none is configured, or swap is full. These failures often also don't have enough warning to allow one to react, whether for a human or for a daemon monitoring PSI.
Here is output from a simple program showing how long it takes in μsec (column 2) to allocate a megabyte of anonymous memory (column 1) when a cgroup is already beyond its memory high setting, and no swap is available:
[root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \
> --wait -t timeout 300 /root/mdf
[...]
95 1035
96 1038
97 1000
98 1036
99 1048
100 1590
101 1968
102 1776
103 1863
104 1757
105 1921
106 1893
107 1760
108 1748
109 1843
110 1716
111 1924
112 1776
113 1831
114 1766
115 1836
116 1588
117 1912
118 1802
119 1857
120 1731
[...]
[System OOM in 2-3 seconds]The delay does go up extremely marginally past the 100MB memory.high threshold, as now we spend time scanning before returning to usermode, but it's nowhere near enough to contain growth. It also doesn't get worse the more pages you have, since it only considers nr_pages.
The current situation goes against both the expectations of users of memory.high, and our intentions as cgroup v2 developers. In cgroup-v2.txt, we claim that we will throttle and only under "extreme conditions" will memory.high protection be breached. Likewise, cgroup v2 users generally also expect that memory.high should throttle workloads as they exceed their high threshold. However, as seen above, this isn't always how it works in practice -- even on banal setups like those with no swap, or where swap has become exhausted, we can end up with memory.high being breached and us having no weapons left in our arsenal to combat runaway growth with, since reclaim is futile.
It's also hard for system monitoring software or users to tell how bad the situation is, as "high" events for the memcg may in some cases be benign, and in others be catastrophic. The current status quo is that we fail containment in a way that doesn't provide any advance warning that things are about to go horribly wrong (for example, we are about to invoke the kernel OOM killer).
This patch introduces explicit throttling when reclaim is failing to keep memcg size contained at the memory.high setting. It does so by applying an exponential delay curve derived from the memcg's overage compared to memory.high. In the normal case where the memcg is either below or only marginally over its memory.high setting, no throttling will be performed.
This composes well with system health monitoring and remediation, as these allocator delays are factored into PSI's memory pressure calculations. This both creates a mechanism system administrators or applications consuming the PSI interface to trivially see that the memcg in question is struggling and use that to make more reasonable decisions, and permits them enough time to act. Either of these can act with significantly more nuance than that we can provide using the system OOM killer.
This is a similar idea to memory.oom_control in cgroup v1 which would put the cgroup to sleep if the threshold was violated, but it's also significantly improved as it results in visible memory pressure, and also doesn't schedule indefinitely, which previously made tracing and other introspection difficult (ie. it's clamped at 2*HZ per allocation through MEMCG_MAX_HIGH_DELAY_JIFFIES).
Contrast the previous results with a kernel with this patch:
[root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \
> --wait -t timeout 300 /root/mdf
[...]
95 1002
96 1000
97 1002
98 1003
99 1000
100 1043
101 84724
102 330628
103 610511
104 1016265
105 1503969
106 2391692
107 2872061
108 3248003
109 4791904
110 5759832
111 6912509
112 8127818
113 9472203
114 12287622
115 12480079
116 14144008
117 15808029
118 16384500
119 16383242
120 16384979
[...]As you can see, in the normal case, memory allocation takes around 1000 μsec. However, as we exceed our memory.high, things start to increase exponentially, but fairly leniently at first. Our first megabyte over memory.high takes us 0.16 seconds, then the next is 0.46 seconds, then the next is almost an entire second. This gets worse until we reach our eventual 2*HZ clamp per batch, resulting in 16 seconds per megabyte. However, this is still making forward progress, so permits tracing or further analysis with programs like GDB.
We use an exponential curve for our delay penalty for a few reasons:
1. We run mem_cgroup_handle_over_high to potentially do reclaim after we've already performed allocations, which means that temporarily going over memory.high by a small amount may be perfectly legitimate, even for compliant workloads. We don't want to unduly penalise such cases.
2. An exponential curve (as opposed to a static or linear delay) allows ramping up memory pressure stats more gradually, which can be useful to work out that you have set memory.high too low, without destroying application performance entirely.
This patch expands on earlier work by Johannes Weiner. Thanks!
--- mm/memcontrol.c | 125 +++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 124 insertions(+), 1 deletion(-)
[v4: Rebased and fixed theoretical (but somewhat unlikely) divide by zero]
mm/memcontrol.c +124 -1
@@ -57,6 +57,7 @@#include <linux/lockdep.h> #include <linux/file.h> #include <linux/tracehook.h> +#include <linux/psi.h> #include <linux/seq_buf.h> #include "internal.h" #include <net/sock.h>
@@ -2314,12 +2315,68 @@ static void high_work_func(struct work_struct *work)reclaim_high (memcg,MEMCG_CHARGE_BATCH ,GFP_KERNEL ); } +/* +* Clamp the maximum sleep time per allocation batch to 2 seconds. This is +* enough to still cause a significant slowdown in most cases, while still +* allowing diagnostics and tracing to proceed without becoming stuck. +*/ +#define MEMCG_MAX_HIGH_DELAY_JIFFIES (2UL*HZ) + +/* +* When calculating the delay, we use these either side of the exponentiation to +* maintain precision and scale to a reasonable number of jiffies (see the table +* below. +* +* - MEMCG_DELAY_PRECISION_SHIFT: Extra precision bits while translating the +* overage ratio to a delay. +* - MEMCG_DELAY_SCALING_SHIFT: The number of bits to scale down down the +* proposed penalty in order to reduce to a reasonable number of jiffies, and +* to produce a reasonable delay curve. +* +* MEMCG_DELAY_SCALING_SHIFT just happens to be a number that produces a +* reasonable delay curve compared to precision-adjusted overage, not +* penalising heavily at first, but still making sure that growth beyond the +* limit penalises misbehaviour cgroups by slowing them down exponentially. For +* example, with a high of 100 megabytes: +* +* +-------+------------------------+ +* | usage | time to allocate in ms | +* +-------+------------------------+ +* | 100M | 0 | +* | 101M | 6 | +* | 102M | 25 | +* | 103M | 57 | +* | 104M | 102 | +* | 105M | 159 | +* | 106M | 230 | +* | 107M | 313 | +* | 108M | 409 | +* | 109M | 518 | +* | 110M | 639 | +* | 111M | 774 | +* | 112M | 921 | +* | 113M | 1081 | +* | 114M | 1254 | +* | 115M | 1439 | +* | 116M | 1638 | +* | 117M | 1849 | +* | 118M | 2000 | +* | 119M | 2000 | +* | 120M | 2000 | +* +-------+------------------------+ +*/ +#define MEMCG_DELAY_PRECISION_SHIFT 20 +#define MEMCG_DELAY_SCALING_SHIFT 14 +/* * Scheduled by try_charge() to be executed from the userland return path * and reclaims memory over the high limit. */ void mem_cgroup_handle_over_high (void ) { +unsigned long usage, high, clamped_high; +unsigned long pflags; +unsigned long penalty_jiffies, overage;unsigned int nr_pages = current->memcg_nr_pages_over_high;struct mem_cgroup *memcg;
@@ -2328,8 +2385,74 @@ void mem_cgroup_handle_over_high(void) memcg =get_mem_cgroup_from_mm (current->mm);reclaim_high (memcg, nr_pages,GFP_KERNEL ); -css_put (&memcg->css); current->memcg_nr_pages_over_high =0 ; + +/* +* memory.high is breached and reclaim is unable to keep up. Throttle +* allocators proactively to slow down excessive growth. +* +* We use overage compared to memory.high to calculate the number of +* jiffies to sleep (penalty_jiffies). Ideally this value should be +* fairly lenient on small overages, and increasingly harsh when the +* memcg in question makes it clear that it has no intention of stopping +* its crazy behaviour, so we exponentially increase the delay based on +* overage amount. +*/ + + usage =page_counter_read (&memcg->memory); + high =READ_ONCE (memcg->high); + +if (usage <= high) + goto out; + +/* +* Prevent division by 0 in overage calculation by acting as if it was a +* threshold of 1 page +*/ + clamped_high =max (high,1 ); + + overage = ((u64)(usage - high) <<MEMCG_DELAY_PRECISION_SHIFT ) + / clamped_high; + penalty_jiffies = ((u64 )overage * overage *HZ ) + >> (MEMCG_DELAY_PRECISION_SHIFT +MEMCG_DELAY_SCALING_SHIFT ); + +/* +* Factor in the task's own contribution to the overage, such that four +* N-sized allocations are throttled approximately the same as one +* 4N-sized allocation. +* +* MEMCG_CHARGE_BATCH pages is nominal, so work out how much smaller or +* larger the current charge patch is than that. +*/ + penalty_jiffies = penalty_jiffies * nr_pages /MEMCG_CHARGE_BATCH ; + +/* +* Clamp the max delay per usermode return so as to still keep the +* application moving forwards and also permit diagnostics, albeit +* extremely slowly. +*/ + penalty_jiffies =min (penalty_jiffies,MEMCG_MAX_HIGH_DELAY_JIFFIES ); + +/* +* Don't sleep if the amount of jiffies this memcg owes us is so low +* that it's not even worth doing, in an attempt to be nice to those who +* go only a small amount over their memory.high value and maybe haven't +* been aggressively reclaimed enough yet. +*/ +if (penalty_jiffies <=HZ /100 ) + goto out; + +/* +* If we exit early, we're guaranteed to die (since +* schedule_timeout_killable sets TASK_KILLABLE). This means we don't +* need to account for any ill-begotten jiffies to pay them off later. +*/ +psi_memstall_enter (&pflags); +schedule_timeout_killable (penalty_jiffies); +psi_memstall_leave (&pflags); + +out: +css_put (&memcg->css); } static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
Signature · 1 line
2.22.0
Quoted 155 lines · Chris Down
We're trying to use memory.high to limit workloads, but have found that containment can frequently fail completely and cause OOM situations outside of the cgroup. This happens especially with swap space -- either when none is configured, or swap is full. These failures often also don't have enough warning to allow one to react, whether for a human or for a daemon monitoring PSI.
Here is output from a simple program showing how long it takes in μsec (column 2) to allocate a megabyte of anonymous memory (column 1) when a cgroup is already beyond its memory high setting, and no swap is available:
[root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \ > --wait -t timeout 300 /root/mdf [...] 95 1035 96 1038 97 1000 98 1036 99 1048 100 1590 101 1968 102 1776 103 1863 104 1757 105 1921 106 1893 107 1760 108 1748 109 1843 110 1716 111 1924 112 1776 113 1831 114 1766 115 1836 116 1588 117 1912 118 1802 119 1857 120 1731 [...] [System OOM in 2-3 seconds]The delay does go up extremely marginally past the 100MB memory.high threshold, as now we spend time scanning before returning to usermode, but it's nowhere near enough to contain growth. It also doesn't get worse the more pages you have, since it only considers
nr_pages.The current situation goes against both the expectations of users of memory.high, and our intentions as cgroup v2 developers. In cgroup-v2.txt, we claim that we will throttle and only under "extreme conditions" will memory.high protection be breached. Likewise, cgroup v2 users generally also expect that memory.high should throttle workloads as they exceed their high threshold. However, as seen above, this isn't always how it works in practice -- even on banal setups like those with no swap, or where swap has become exhausted, we can end up with memory.high being breached and us having no weapons left in our arsenal to combat runaway growth with, since reclaim is futile.
It's also hard for system monitoring software or users to tell how bad the situation is, as "high" events for the memcg may in some cases be benign, and in others be catastrophic. The current status quo is that we fail containment in a way that doesn't provide any advance warning that things are about to go horribly wrong (for example, we are about to invoke the kernel OOM killer).
This patch introduces explicit throttling when reclaim is failing to keep memcg size contained at the memory.high setting. It does so by applying an exponential delay curve derived from the memcg's overage compared to memory.high. In the normal case where the memcg is either below or only marginally over its memory.high setting, no throttling will be performed.
This composes well with system health monitoring and remediation, as these allocator delays are factored into PSI's memory pressure calculations. This both creates a mechanism system administrators or applications consuming the PSI interface to trivially see that the memcg in question is struggling and use that to make more reasonable decisions, and permits them enough time to act. Either of these can act with significantly more nuance than that we can provide using the system OOM killer.
This is a similar idea to memory.oom_control in cgroup v1 which would put the cgroup to sleep if the threshold was violated, but it's also significantly improved as it results in visible memory pressure, and also doesn't schedule indefinitely, which previously made tracing and other introspection difficult (ie. it's clamped at 2*HZ per allocation through
MEMCG_MAX_HIGH_DELAY_JIFFIES).Contrast the previous results with a kernel with this patch:
[root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \ > --wait -t timeout 300 /root/mdf [...] 95 1002 96 1000 97 1002 98 1003 99 1000 100 1043 101 84724 102 330628 103 610511 104 1016265 105 1503969 106 2391692 107 2872061 108 3248003 109 4791904 110 5759832 111 6912509 112 8127818 113 9472203 114 12287622 115 12480079 116 14144008 117 15808029 118 16384500 119 16383242 120 16384979 [...]As you can see, in the normal case, memory allocation takes around 1000 μsec. However, as we exceed our memory.high, things start to increase exponentially, but fairly leniently at first. Our first megabyte over memory.high takes us 0.16 seconds, then the next is 0.46 seconds, then the next is almost an entire second. This gets worse until we reach our eventual 2*HZ clamp per batch, resulting in 16 seconds per megabyte. However, this is still making forward progress, so permits tracing or further analysis with programs like GDB.
We use an exponential curve for our delay penalty for a few reasons:
1. We run
mem_cgroup_handle_over_highto potentially do reclaim after we've already performed allocations, which means that temporarily going over memory.high by a small amount may be perfectly legitimate, even for compliant workloads. We don't want to unduly penalise such cases.
2. An exponential curve (as opposed to a static or linear delay) allows ramping up memory pressure stats more gradually, which can be useful to work out that you have set memory.high too low, without destroying application performance entirely.This patch expands on earlier work by Johannes Weiner. Thanks!
Signed-off-by Chris Down <[email protected]>Cc Andrew Morton <[email protected]>Cc Johannes Weiner <[email protected]>Cc Tejun Heo <[email protected]>Cc Roman Gushchin <[email protected]>
Quoted 1 line
---