189 lines · plain
1.. _psi:2 3================================4PSI - Pressure Stall Information5================================6 7:Date: April, 20188:Author: Johannes Weiner <hannes@cmpxchg.org>9 10When CPU, memory or IO devices are contended, workloads experience11latency spikes, throughput losses, and run the risk of OOM kills.12 13Without an accurate measure of such contention, users are forced to14either play it safe and under-utilize their hardware resources, or15roll the dice and frequently suffer the disruptions resulting from16excessive overcommit.17 18The psi feature identifies and quantifies the disruptions caused by19such resource crunches and the time impact it has on complex workloads20or even entire systems.21 22Having an accurate measure of productivity losses caused by resource23scarcity aids users in sizing workloads to hardware--or provisioning24hardware according to workload demand.25 26As psi aggregates this information in realtime, systems can be managed27dynamically using techniques such as load shedding, migrating jobs to28other systems or data centers, or strategically pausing or killing low29priority or restartable batch jobs.30 31This allows maximizing hardware utilization without sacrificing32workload health or risking major disruptions such as OOM kills.33 34Pressure interface35==================36 37Pressure information for each resource is exported through the38respective file in /proc/pressure/ -- cpu, memory, and io.39 40The format is as such::41 42 some avg10=0.00 avg60=0.00 avg300=0.00 total=043 full avg10=0.00 avg60=0.00 avg300=0.00 total=044 45The "some" line indicates the share of time in which at least some46tasks are stalled on a given resource.47 48The "full" line indicates the share of time in which all non-idle49tasks are stalled on a given resource simultaneously. In this state50actual CPU cycles are going to waste, and a workload that spends51extended time in this state is considered to be thrashing. This has52severe impact on performance, and it's useful to distinguish this53situation from a state where some tasks are stalled but the CPU is54still doing productive work. As such, time spent in this subset of the55stall state is tracked separately and exported in the "full" averages.56 57CPU full is undefined at the system level, but has been reported58since 5.13, so it is set to zero for backward compatibility.59 60The ratios (in %) are tracked as recent trends over ten, sixty, and61three hundred second windows, which gives insight into short term events62as well as medium and long term trends. The total absolute stall time63(in us) is tracked and exported as well, to allow detection of latency64spikes which wouldn't necessarily make a dent in the time averages,65or to average trends over custom time frames.66 67Monitoring for pressure thresholds68==================================69 70Users can register triggers and use poll() to be woken up when resource71pressure exceeds certain thresholds.72 73A trigger describes the maximum cumulative stall time over a specific74time window, e.g. 100ms of total stall time within any 500ms window to75generate a wakeup event.76 77To register a trigger user has to open psi interface file under78/proc/pressure/ representing the resource to be monitored and write the79desired threshold and time window. The open file descriptor should be80used to wait for trigger events using select(), poll() or epoll().81The following format is used::82 83 <some|full> <stall amount in us> <time window in us>84 85For example writing "some 150000 1000000" into /proc/pressure/memory86would add 150ms threshold for partial memory stall measured within871sec time window. Writing "full 50000 1000000" into /proc/pressure/io88would add 50ms threshold for full io stall measured within 1sec time window.89 90Triggers can be set on more than one psi metric and more than one trigger91for the same psi metric can be specified. However for each trigger a separate92file descriptor is required to be able to poll it separately from others,93therefore for each trigger a separate open() syscall should be made even94when opening the same psi interface file. Write operations to a file descriptor95with an already existing psi trigger will fail with EBUSY.96 97Monitors activate only when system enters stall state for the monitored98psi metric and deactivates upon exit from the stall state. While system is99in the stall state psi signal growth is monitored at a rate of 10 times per100tracking window.101 102The kernel accepts window sizes ranging from 500ms to 10s, therefore min103monitoring update interval is 50ms and max is 1s. Min limit is set to104prevent overly frequent polling. Max limit is chosen as a high enough number105after which monitors are most likely not needed and psi averages can be used106instead.107 108Unprivileged users can also create monitors, with the only limitation that the109window size must be a multiple of 2s, in order to prevent excessive resource110usage.111 112When activated, psi monitor stays active for at least the duration of one113tracking window to avoid repeated activations/deactivations when system is114bouncing in and out of the stall state.115 116Notifications to the userspace are rate-limited to one per tracking window.117 118The trigger will de-register when the file descriptor used to define the119trigger is closed.120 121Userspace monitor usage example122===============================123 124::125 126 #include <errno.h>127 #include <fcntl.h>128 #include <stdio.h>129 #include <poll.h>130 #include <string.h>131 #include <unistd.h>132 133 /*134 * Monitor memory partial stall with 1s tracking window size135 * and 150ms threshold.136 */137 int main() {138 const char trig[] = "some 150000 1000000";139 struct pollfd fds;140 int n;141 142 fds.fd = open("/proc/pressure/memory", O_RDWR | O_NONBLOCK);143 if (fds.fd < 0) {144 printf("/proc/pressure/memory open error: %s\n",145 strerror(errno));146 return 1;147 }148 fds.events = POLLPRI;149 150 if (write(fds.fd, trig, strlen(trig) + 1) < 0) {151 printf("/proc/pressure/memory write error: %s\n",152 strerror(errno));153 return 1;154 }155 156 printf("waiting for events...\n");157 while (1) {158 n = poll(&fds, 1, -1);159 if (n < 0) {160 printf("poll error: %s\n", strerror(errno));161 return 1;162 }163 if (fds.revents & POLLERR) {164 printf("got POLLERR, event source is gone\n");165 return 0;166 }167 if (fds.revents & POLLPRI) {168 printf("event triggered!\n");169 } else {170 printf("unknown event received: 0x%x\n", fds.revents);171 return 1;172 }173 }174 175 return 0;176 }177 178Cgroup2 interface179=================180 181In a system with a CONFIG_CGROUPS=y kernel and the cgroup2 filesystem182mounted, pressure stall information is also tracked for tasks grouped183into cgroups. Each subdirectory in the cgroupfs mountpoint contains184cpu.pressure, memory.pressure, and io.pressure files; the format is185the same as the /proc/pressure/ files.186 187Per-cgroup psi monitors can be specified and used the same way as188system-wide ones.189