Container Performance Analysis Brendan Gregg, Netflix

1. Container Performance Analysis Brendan Gregg Sr. Performance Architect, Netflix

2. Identify bottlenecks: 1.  In the host vs container, using system metrics 2.  In application code on containers, using CPU flame graphs 3.  Deeper in the kernel, using tracing tools Focus of this talk is how containers work in Linux (will demo on 4.9) I will include some Docker specifics, and start with a Netflix summary (Titus) Take Aways

3. Containers at 1. Titus Summary slides from the Titus team

4. Titus •  Cloud runtime platform for container jobs •  Scheduling •  Service & batch job management •  Advanced resource management across elastic shared resource pool •  Container Execution •  Docker and AWS EC2 Integration •  Adds VPC, security groups, EC2 metadata, IAM roles, S3 logs, … •  Integration with Netflix infrastructure •  In depth: http://techblog.netflix.com/2017/04/the-evolution-of-container-usage-at.html Service Job Management Resource Management & Op=miza=on Container Execu=on Integra=on Batch

5. Current Titus Scale •  Deployed across multiple AWS accounts & three regions •  Over 2,500 instances (Mostly M4.4xls & R3.8xls) •  Over a week period launched over 1,000,000 containers

6. Titus Use Cases •  Service •  Stream Processing (Flink) •  UI Services (Node.JS single core) •  Internal dashboards •  Batch •  Algorithm training, personalization & recommendations •  Adhoc reporting •  Continuous integration builds •  Queued worker model •  Media encoding

7. Container Performance @Netflix •  Ability to scale and balance workloads with EC2 and Titus •  Can already solve many perf issues •  Performance needs: •  Application analysis: using CPU flame graphs with containers •  Host tuning: file system, networking, sysctl's, … •  Container analysis and tuning: cgroups, GPUs, … •  Capacity planning: reduce over provisioning

8. And Strategy 2. Container Background

9. Namespaces Kernel PID namespace 1 Host PID 1 1237 1 (1238) 2 (1241) … Namespaces: •  cgroup •  ipc •  mnt •  net •  pid •  user •  uts Restricting visibility PID namespaces

10. Control Groups CPUs container 1 cgroups: •  blkio •  cpu,cpuacct •  cpuset •  devices •  hugetlb •  memory •  net_cls,net_prio •  pids •  … Restricting usage CPU cgroups container 2 container 3 cpu cgroup 1 … 2 … 3 …

11. Linux Containers Container 1 (namespaces) Container = combination of namespaces & cgroups Host … cgroups Kernel cgroups cgroups Container 2 (namespaces) Container 3 (namespaces)

12. cgroup v1 cpu,cpuacct: •  cap CPU usage (hard limit). e.g. 1.5 CPUs. •  CPU shares. e.g. 100 shares. •  usage statistics (cpuacct) memory: •  limit and kmem limit (maximum bytes) •  OOM control: enable/disable •  usage statistics blkio (block I/O): •  weights (like shares) •  IOPS/tput caps per storage device •  statistics Docker: --cpus (1.13) --cpu-shares --memory --kernel-memory --oom-kill-disable

13. container's shares total busy shares CPU Shares Container's CPU limit = 100% x This lets a container use other tenant's idle CPU (aka "bursting"), when available. container's shares total allocated shares Container's minimum CPU limit = 100% x Can make analysis tricky. Why did perf regress? Less bursting available?

14. •  Major rewrite has been happening: cgroups v2 •  Supports nested groups, better organization and consistency •  Some already merged, some not yet (e.g. CPU) •  See docs/talks by maintainer Tejun Heo (Facebook) •  References: •  https://www.kernel.org/doc/Documentation/cgroup-v2.txt •  https://lwn.net/Articles/679786/ cgroup v2

15. File systems •  Containers may be setup with aufs/overlay on top of another FS •  See "in practice" pages and their performance sections from https://docs.docker.com/engine/userguide/storagedriver/ Networking •  With Docker, can be bridge, host, or overlay networks •  Overlay networks have come with significant performance cost Container OS Configuration

16. Performance analysis with containers: •  One kernel •  Two perspectives •  Namespaces •  cgroups Methodologies: •  USE Method •  Workload characterization •  Checklists •  Event tracing Analysis Strategy

17. USE Method For every resource, check: 1.  Utilization 2.  Saturation 3.  Errors For example, CPUs: •  Utilization: time busy •  Saturation: run queue length or latency •  Errors: ECC errors, etc. Can be applied to hardware resources and software resources (cgroups) Resource Utilization (%)X

18. And Container Awareness … if you have host access 3. Host Tools

19. •  PIDs in host don't match those seen in containers •  Symbol files aren't where tools expect them •  The kernel currently doesn't have a container ID Host Analysis Challenges

20. I'll demo CLI tools It's the lowest common denominator You may usually use GUIs (like we do). They source the same metrics. CLI Tool Disclaimer

21. 3.1. Host Physical Resources A refresher of basics... Not container specific. This will, however, solve many issues! Containers are often not the problem.

22. Linux Perf Tools Where can we begin?

23. Host Perf Analysis in 60s http://techblog.netflix.com/2015/11/linux-performance-analysis-in-60s.html 1.  uptime 2.  dmesg | tail 3.  vmstat 1 4.  mpstat -P ALL 1 5.  pidstat 1 6.  iostat -xz 1 7.  free -m 8.  sar -n DEV 1 9.  sar -n TCP,ETCP 1 10.  top load averages kernel errors overall stats by time CPU balance process usage disk I/O memory usage network I/O TCP stats check overview

24. USE Method: Host Resources Resource Utilization Saturation Errors CPU mpstat -P ALL 1, sum non-idle fields vmstat 1, "r" perf Memory Capacity free –m, "used"/"total" vmstat 1, "si"+"so"; demsg | grep killed dmesg Storage I/O iostat –xz 1, "%util" iostat –xnz 1, "avgqu-sz" > 1 /sys/…/ioerr_cnt; smartctl Network nicstat, "%Util" ifconfig, "overrunns"; netstat –s "retrans…" ifconfig, "errors" These should be in your monitoring GUI. Can do other resources too (busses, ...)

25. Event Tracing: e.g. iosnoop Disk I/O events with latency (from perf-tools; also in bcc/BPF as biosnoop) # ./iosnoop Tracing block I/O... Ctrl-C to end. COMM PID TYPE DEV BLOCK BYTES LATms supervise 1809 W 202,1 17039968 4096 1.32 supervise 1809 W 202,1 17039976 4096 1.30 tar 14794 RM 202,1 8457608 4096 7.53 tar 14794 RM 202,1 8470336 4096 14.90 tar 14794 RM 202,1 8470368 4096 0.27 tar 14794 RM 202,1 8470784 4096 7.74 tar 14794 RM 202,1 8470360 4096 0.25 tar 14794 RM 202,1 8469968 4096 0.24 tar 14794 RM 202,1 8470240 4096 0.24 tar 14794 RM 202,1 8470392 4096 0.23

26. Event Tracing: e.g. zfsslower •  This is from our production Titan system (Docker). •  File system latency is a better pain indicator than disk latency. •  zfsslower (and btrfs*, etc) are in bcc/BPF. Can exonerate FS/disks. # /usr/share/bcc/tools/zfsslower 1 Tracing ZFS operations slower than 1 ms TIME COMM PID T BYTES OFF_KB LAT(ms) FILENAME 23:44:40 java 31386 O 0 0 8.02 solrFeatures.txt 23:44:53 java 31386 W 8190 1812222 36.24 solrFeatures.txt 23:44:59 java 31386 W 8192 1826302 20.28 solrFeatures.txt 23:44:59 java 31386 W 8191 1826846 28.15 solrFeatures.txt 23:45:00 java 31386 W 8192 1831015 32.17 solrFeatures.txt 23:45:15 java 31386 O 0 0 27.44 solrFeatures.txt 23:45:56 dockerd 3599 S 0 0 1.03 .tmp-a66ce9aad… 23:46:16 java 31386 W 31 0 36.28 solrFeatures.txt

27. Latency Histograms: e.g. btrfsdist # ./btrfsdist Tracing btrfs operation latency... Hit Ctrl-C to end. ^C operation = 'read' usecs : count distribution 0 -> 1 : 192529 |****************************************| 2 -> 3 : 72337 |*************** | 4 -> 7 : 5620 |* | 8 -> 15 : 1026 | | 16 -> 31 : 369 | | 32 -> 63 : 239 | | 64 -> 127 : 53 | | 128 -> 255 : 975 | | 256 -> 511 : 524 | | 512 -> 1023 : 128 | | 1024 -> 2047 : 16 | | 2048 -> 4095 : 7 | | 4096 -> 8191 : 2 | | probably cache reads probably cache misses (flash reads)

28. Latency Histograms: e.g. btrfsdist •  From a test Titan system (Docker). •  Histograms show modes, outliers. Also in bcc/BPF (with other FSes). •  Latency heat maps: http://queue.acm.org/detail.cfm?id=1809426 […] operation = 'write' usecs : count distribution 0 -> 1 : 1 | | 2 -> 3 : 276 | | 4 -> 7 : 32125 |*********** | 8 -> 15 : 111253 |****************************************| 16 -> 31 : 59154 |********************* | 32 -> 63 : 5463 |* | 64 -> 127 : 612 | | 128 -> 255 : 25 | | 256 -> 511 : 2 | | 512 -> 1023 : 1 | |

29. 3.2. Host Containers & cgroups Inspecting containers from the host

30. Namespaces # ./dockerpsns.sh CONTAINER NAME PID PATH CGROUP IPC MNT NET PID USER UTS host titusagent-mainvpc-m 1 systemd 4026531835 4026531839 4026531840 4026532533 4026531836 4026531837 4026531838 b27909cd6dd1 Titus-1435830-worker 37280 svscanboot 4026531835 4026533387 4026533385 4026532931 4026533388 4026531837 4026533386 dcf3a506de45 Titus-1392192-worker 27992 /apps/spaas/spaa 4026531835 4026533354 4026533352 4026532991 4026533355 4026531837 4026533353 370a3f041f36 Titus-1243558-worker 98602 /apps/spaas/spaa 4026531835 4026533290 4026533288 4026533223 4026533291 4026531837 4026533289 af7549c76d9a Titus-1243553-worker 97972 /apps/spaas/spaa 4026531835 4026533216 4026533214 4026533149 4026533217 4026531837 4026533215 dc27769a9b9c Titus-1243546-worker 97356 /apps/spaas/spaa 4026531835 4026533142 4026533140 4026533075 4026533143 4026531837 4026533141 e18bd6189dcd Titus-1243517-worker 96733 /apps/spaas/spaa 4026531835 4026533068 4026533066 4026533001 4026533069 4026531837 4026533067 ab45227dcea9 Titus-1243516-worker 96173 /apps/spaas/spaa 4026531835 4026532920 4026532918 4026532830 4026532921 4026531837 4026532919 Worth checking namespace config before analysis: •  A POC "docker ps --namespaces" tool. NS shared with root in red. •  https://github.com/docker/docker/issues/32501

31. systemd-cgtop # systemd-cgtop Control Group Tasks %CPU Memory Input/s Output/s / - 798.2 45.9G - - /docker 1082 790.1 42.1G - - /docker/dcf3a...9d28fc4a1c72bbaff4a24834 200 610.5 24.0G - - /docker/370a3...e64ca01198f1e843ade7ce21 170 174.0 3.0G - - /system.slice 748 5.3 4.1G - - /system.slice/daemontools.service 422 4.0 2.8G - - /docker/dc277...42ab0603bbda2ac8af67996b 160 2.5 2.3G - - /user.slice 5 2.0 34.5M - - /user.slice/user-0.slice 5 2.0 15.7M - - /user.slice/u....slice/session-c26.scope 3 2.0 13.3M - - /docker/ab452...c946f8447f2a4184f3ccff2a 174 1.0 6.3G - - /docker/e18bd...26ffdd7368b870aa3d1deb7a 156 0.8 2.9G - - [...] A "top" for cgroups:

32. docker stats # docker stats CONTAINER CPU % MEM USAGE / LIMIT MEM % NET I/O BLOCK I/O PIDS 353426a09db1 526.81% 4.061 GiB / 8.5 GiB 47.78% 0 B / 0 B 2.818 MB / 0 B 247 6bf166a66e08 303.82% 3.448 GiB / 8.5 GiB 40.57% 0 B / 0 B 2.032 MB / 0 B 267 58dcf8aed0a7 41.01% 1.322 GiB / 2.5 GiB 52.89% 0 B / 0 B 0 B / 0 B 229 61061566ffe5 85.92% 220.9 MiB / 3.023 GiB 7.14% 0 B / 0 B 43.4 MB / 0 B 61 bdc721460293 2.69% 1.204 GiB / 3.906 GiB 30.82% 0 B / 0 B 4.35 MB / 0 B 66 6c80ed61ae63 477.45% 557.7 MiB / 8 GiB 6.81% 0 B / 0 B 9.257 MB / 0 B 19 337292fb5b64 89.05% 766.2 MiB / 8 GiB 9.35% 0 B / 0 B 5.493 MB / 0 B 19 b652ede9a605 173.50% 689.2 MiB / 8 GiB 8.41% 0 B / 0 B 6.48 MB / 0 B 19 d7cd2599291f 504.28% 673.2 MiB / 8 GiB 8.22% 0 B / 0 B 12.58 MB / 0 B 19 05bf9f3e0d13 314.46% 711.6 MiB / 8 GiB 8.69% 0 B / 0 B 7.942 MB / 0 B 19 09082f005755 142.04% 693.9 MiB / 8 GiB 8.47% 0 B / 0 B 8.081 MB / 0 B 19 bd45a3e1ce16 190.26% 538.3 MiB / 8 GiB 6.57% 0 B / 0 B 10.6 MB / 0 B 19 [...] A "top" for containers. Resource utilization. Workload characterization. Loris Degioanni demoed a similar sysdigcloud view yesterday (needs the sysdig kernel agent)

33. top # top - 22:46:53 up 36 days, 59 min, 1 user, load average: 5.77, 5.61, 5.63 Tasks: 1067 total, 1 running, 1046 sleeping, 0 stopped, 20 zombie %Cpu(s): 34.8 us, 1.8 sy, 0.0 ni, 61.3 id, 0.0 wa, 0.0 hi, 1.9 si, 0.1 st KiB Mem : 65958552 total, 12418448 free, 49247988 used, 4292116 buff/cache KiB Swap: 0 total, 0 free, 0 used. 13101316 avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 28321 root 20 0 33.126g 0.023t 37564 S 621.1 38.2 35184:09 java 97712 root 20 0 11.445g 2.333g 37084 S 3.1 3.7 404:27.90 java 98306 root 20 0 12.149g 3.060g 36996 S 2.0 4.9 194:21.10 java 96511 root 20 0 15.567g 6.313g 37112 S 1.7 10.0 168:07.44 java 5283 root 20 0 1643676 100092 94184 S 1.0 0.2 401:36.16 mesos-slave 2079 root 20 0 9512 132 12 S 0.7 0.0 220:07.75 rngd 5272 titusag+ 20 0 10.473g 1.611g 23488 S 0.7 2.6 1934:44 java […] In the host, top shows all processes. Currently doesn't show a container ID. … remember, there is no container ID in the kernel yet.

34. htop CGROUP PID USER PRI NI VIRT RES SHR S CPU% MEM% TIME+ Command :pids:/docker/ 28321 root 20 0 33.1G 24.0G 37564 S 524. 38.2 672h /apps/java :pids:/docker/ 9982 root 20 0 33.1G 24.0G 37564 S 44.4 38.2 17h00:41 /apps/java :pids:/docker/ 9985 root 20 0 33.1G 24.0G 37564 R 41.9 38.2 16h44:51 /apps/java :pids:/docker/ 9979 root 20 0 33.1G 24.0G 37564 S 41.2 38.2 17h01:35 /apps/java :pids:/docker/ 9980 root 20 0 33.1G 24.0G 37564 S 39.3 38.2 16h59:17 /apps/java :pids:/docker/ 9981 root 20 0 33.1G 24.0G 37564 S 39.3 38.2 17h01:32 /apps/java :pids:/docker/ 9984 root 20 0 33.1G 24.0G 37564 S 37.3 38.2 16h49:03 /apps/java :pids:/docker/ 9983 root 20 0 33.1G 24.0G 37564 R 35.4 38.2 16h54:31 /apps/java :pids:/docker/ 9986 root 20 0 33.1G 24.0G 37564 S 35.4 38.2 17h05:30 /apps/java :name=systemd:/user.slice/user-0.slice/session-c31.scope? 74066 root 20 0 27620 :pids:/docker/ 9998 root 20 0 33.1G 24.0G 37564 R 28.3 38.2 11h38:03 /apps/java :pids:/docker/ 10001 root 20 0 33.1G 24.0G 37564 S 27.7 38.2 11h38:59 /apps/java :name=systemd:/system.slice/daemontools.service? 5272 titusagen 20 0 10.5G 1650M 23 :pids:/docker/ 10002 root 20 0 33.1G 24.0G 37564 S 25.1 38.2 11h40:37 /apps/java htop can add a CGROUP field, but, can truncate important info: Can fix, but that would be Docker + cgroup-v1 specific. Still need a kernel CID.

35. Host PID -> Container ID # grep 28321 /sys/fs/cgroup/cpu,cpuacct/docker/*/tasks | cut -d/ -f7 dcf3a506de453107715362f6c9ba9056fcfc6e769d28fc4a1c72bbaff4a24834 … who does that (CPU busy) PID 28321 belong to? •  Only works for Docker, and that cgroup v1 layout. Some Linux commands: # ls -l /proc/27992/ns/* lrwxrwxrwx 1 root root 0 Apr 13 20:49 cgroup -> cgroup:[4026531835] lrwxrwxrwx 1 root root 0 Apr 13 20:49 ipc -> ipc:[4026533354] lrwxrwxrwx 1 root root 0 Apr 13 20:49 mnt -> mnt:[4026533352] […] # cat /proc/27992/cgroup 11:freezer:/docker/dcf3a506de453107715362f6c9ba9056fcfc6e769d28fc4a1c72bbaff4a24834 10:blkio:/docker/dcf3a506de453107715362f6c9ba9056fcfc6e769d28fc4a1c72bbaff4a24834 9:perf_event:/docker/dcf3a506de453107715362f6c9ba9056fcfc6e769d28fc4a1c72bbaff4a24834 […]

36. nsenter Wrapping # nsenter -t 28321 -u hostname titus-1392192-worker-14-16 … what hostname is PID 28321 running on? •  Can namespace enter: •  -m: mount -u: uts -i: ipc -n: net -p: pid -U: user •  Bypasses cgroup limits, and seccomp profile (allowing syscalls) •  For Docker, you can enter the container more completely with: docker exec -it CID command •  Handy nsenter one-liners: •  nsenter -t PID -u hostname container hostname •  nsenter -t PID -n netstat -i container netstat •  nsenter -t PID –m -p df -h container file system usage •  nsenter -t PID -p top container top

37. nsenter: Host -> Container top # nsenter -t 28321 -m -p top top - 18:16:13 up 36 days, 20:28, 0 users, load average: 5.66, 5.29, 5.28 Tasks: 6 total, 1 running, 5 sleeping, 0 stopped, 0 zombie %Cpu(s): 30.5 us, 1.7 sy, 0.0 ni, 65.9 id, 0.0 wa, 0.0 hi, 1.8 si, 0.1 st KiB Mem: 65958552 total, 54664124 used, 11294428 free, 164232 buffers KiB Swap: 0 total, 0 used, 0 free. 1592372 cached Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 301 root 20 0 33.127g 0.023t 37564 S 537.3 38.2 40269:41 java 1 root 20 0 21404 2236 1812 S 0.0 0.0 4:15.11 bash 87888 root 20 0 21464 1720 1348 R 0.0 0.0 0:00.00 top … Given PID 28321, running top for its container by entering its namespaces: Note that it is PID 301 in the container. Can also see this using: # grep NSpid /proc/28321/status NSpid: 28321 301

38. perf: CPU Profiling # perf record -F 49 -a -g -- sleep 30 # perf script Failed to open /lib/x86_64-linux-gnu/libc-2.19.so, continuing without symbols Failed to open /tmp/perf-28321.map, continuing without symbols Can run system-wide (-a), match a pid (-p), or cgroup (-G, if it works) •  Current symbol translation gotchas (up to 4.10-ish): •  perf can't find /tmp/perf-PID.map files in the host, and the PID is different •  perf can't find container binaries under host paths (what /usr/bin/java?) •  Can copy files to the host, map PIDs, then run perf script/report: •  http://blog.alicegoldfuss.com/making-flamegraphs-with-containerized-java/ •  http://batey.info/docker-jvm-flamegraphs.html •  Can nsenter (-m -u -i -n -p) a "power" shell, and then run "perf -p PID" •  perf should be fixed to be namespace aware (like bcc was, PR#1051)

39. •  See previous slide for getting perf symbols to work •  From the host, can study all containers, as well as container overheads CPU Flame Graphs git clone --depth 1 https://github.com/brendangregg/FlameGraph cd FlameGraph perf record –F 49 -a –g -- sleep 30 perf script | ./stackcollapse-perf.pl | ./flamegraph.pl > perf.svg Java, missing stacks (need -XX:+PreserveFramePointer) Kernel TCP/IP stack Look in areas like this to find and quantify overhead (cgroup throttles, FS layers, networking, etc). It's likely small and hard to find.

40. /sys/fs/cgroups (raw) # cd /sys/fs/cgroup/cpu,cpuacct/docker/02a7cf65f82e3f3e75283944caa4462e82f8f6ff5a7c9a... # ls cgroup.clone_children cpuacct.usage_all cpuacct.usage_sys cpu.shares cgroup.procs cpuacct.usage_percpu cpuacct.usage_user cpu.stat cpuacct.stat cpuacct.usage_percpu_sys cpu.cfs_period_us notify_on_release cpuacct.usage cpuacct.usage_percpu_user cpu.cfs_quota_us tasks # cat cpuacct.usage 1615816262506 # cat cpu.stat nr_periods 507 nr_throttled 74 throttled_time 3816445175 The best source for per-cgroup metrics. e.g. CPU: •  https://www.kernel.org/doc/Documentation/cgroup-v1/, ../scheduler/sched-bwc.txt •  https://blog.docker.com/2013/10/gathering-lxc-docker-containers-metrics/ Note: grep cgroup /proc/mounts to check where these are mounted These metrics should be included in performance monitoring GUIs total time throttled (nanoseconds). saturation metric. average throttle time = throttled_time / nr_throttled

41. Cloud-wide monitoring of containers (and instances) Fetches cgroup metrics via Intel snap https://github.com/netflix/Atlas Netflix Atlas

42. Our per-instance analyzer Has per-container metrics https://github.com/Netflix/vector Netflix Vector

43. A metric collector used by monitoring GUIs https://github.com/intelsdi-x/snap Has a Docker plugin to read cgroup stats There's also a collectd plugin: https://github.com/bobrik/collectd- docker Intel snap

44. 3.3. Let's Play a Game Host or Container? (or neither?)

45. Game Scenario 1 Container user claims they have a CPU performance issue •  Container has a CPU cap and CPU shares configured •  There is idle CPU on the host •  Other tenants are CPU busy •  /sys/fs/cgroup/.../cpu.stat -> throttled_time is increasing •  /proc/PID/status nonvoluntary_ctxt_switches is increasing •  Container CPU usage equals its cap (clue: this is not really a clue)

46. Game Scenario 2 Container user claims they have a CPU performance issue •  Container has a CPU cap and CPU shares configured •  There is no idle CPU on the host •  Other tenants are CPU busy •  /sys/fs/cgroup/.../cpu.stat -> throttled_time is not increasing •  /proc/PID/status nonvoluntary_ctxt_switches is increasing

47. Game Scenario 3 Container user claims they have a CPU performance issue •  Container has CPU shares configured •  There is no idle CPU on the host •  Other tenants are CPU busy •  /sys/fs/cgroup/.../cpu.stat -> throttled_time is not increasing •  /proc/PID/status nonvoluntary_ctxt_switches is not increasing much Experiments to confirm conclusion?

48. Methodology: Reverse Diagnosis Enumerate possible outcomes, and work backwards to the metrics needed for diagnosis. e.g. CPU performance outcomes: A.  physical CPU throttled B.  cap throttled C.  shares throttled (assumes physical CPU limited as well) D.  not throttled Gameanswers:1.B,2.C,3.D

49. throttled_time increasing? cap throttled Y nonvol…switches increasing? not throttled N Y host has idle CPU? Y (but dig further) physical CPU throttledshare throttled all other tenants idle? N Y N N CPU Bottleneck Identification

50. And Container Awareness … if you only have guest access 4. Guest Tools

51. •  Some resource metrics are for the container, some for the host. Confusing! •  May lack system capabilities or syscalls to run profilers and tracers Guest Analysis Challenges

52. CPU container# uptime 20:17:19 up 45 days, 21:21, 0 users, load average: 5.08, 3.69, 2.22 container# mpstat 1 Linux 4.9.0 (02a7cf65f82e) 04/14/17 _x86_64_ (8 CPU) 20:17:26 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %gnice %idle 20:17:27 all 51.00 0.00 12.28 0.00 0.00 0.00 0.00 0.00 0.00 36.72 20:17:28 all 50.88 0.00 12.31 0.00 0.00 0.00 0.00 0.00 0.00 36.81 ^C Average: all 50.94 0.00 12.30 0.00 0.00 0.00 0.00 0.00 0.00 36.76 container# pidstat 1 Linux 4.9.0 (02a7cf65f82e) 04/14/17 _x86_64_ (8 CPU) 20:17:33 UID PID %usr %system %guest %CPU CPU Command 20:17:34 UID PID %usr %system %guest %CPU CPU Command 20:17:35 UID PID %usr %system %guest %CPU CPU Command [...] Can see host's CPU devices, but only container (pid namespace) processes: load! busy CPUs but this container is running nothing (we saw CPU usage from neighbors)

53. Memory container# free -m total used free shared buff/cache available Mem: 15040 1019 8381 153 5639 14155 Swap: 0 0 0 container# perl -e '$a = "A" x 1_000_000_000' Killed Can see host's memory: tries to consume ~2 Gbytes host memory (this container is --memory=1g)

54. Disks container# iostat -xz 1 avg-cpu: %user %nice %system %iowait %steal %idle 52.57 0.00 16.94 0.00 0.00 30.49 Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s avgrq-sz avgqu-sz await r_await w_await svctm %util xvdap1 0.00 7.00 0.00 2.00 0.00 36.00 36.00 0.00 2.00 0.00 2.00 2.00 0.40 xvdb 0.00 0.00 200.00 0.00 3080.00 0.00 30.80 0.04 0.20 0.20 0.00 0.20 4.00 xvdc 0.00 0.00 185.00 0.00 2840.00 0.00 30.70 0.04 0.24 0.24 0.00 0.24 4.40 md0 0.00 0.00 385.00 0.00 5920.00 0.00 30.75 0.00 0.00 0.00 0.00 0.00 0.00 [...] container# pidstat -d 1 Linux 4.9.0 (02a7cf65f82e) 04/18/17 _x86_64_ (8 CPU) 22:41:13 UID PID kB_rd/s kB_wr/s kB_ccwr/s iodelay Command 22:41:14 UID PID kB_rd/s kB_wr/s kB_ccwr/s iodelay Command 22:41:15 UID PID kB_rd/s kB_wr/s kB_ccwr/s iodelay Command [...] Can see host's disk devices: host disk I/O but no container I/O

55. Network container# sar -n DEV,TCP 1 Linux 4.9.0 (02a7cf65f82e) 04/14/17 _x86_64_ (8 CPU) 21:45:07 IFACE rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s %ifutil 21:45:08 lo 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 21:45:08 eth0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 21:45:07 active/s passive/s iseg/s oseg/s 21:45:08 0.00 0.00 0.00 0.00 21:45:08 IFACE rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s %ifutil 21:45:09 lo 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 21:45:09 eth0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 21:45:08 active/s passive/s iseg/s oseg/s 21:45:09 0.00 0.00 0.00 0.00 [...] Can't see host's network interfaces (network namespace): host has heavy network I/O, container sees itself (idle)

56. This confuses apps too: trying to bind on all CPUs, or using 25% of memory •  Including the JDK, which is unaware of container limits, covered yesterday by Fabiane Nardon We could add a "metrics" namespace so the container only sees itself •  Or enhance existing namespaces to do this If you add a metrics namespace, please consider adding an option for: •  /proc/host/stats: maps to host's /proc/stats, for CPU stats •  /proc/host/diskstats: maps to host's /proc/diskstats, for disk stats As those host metrics can be useful, to identify/exonerate neighbor issues Metrics Namespace

57. perf: CPU Profiling container# ./perf record -F 99 -a -g -- sleep 10 perf_event_open(..., PERF_FLAG_FD_CLOEXEC) failed with unexpected error 1 (Operation not permitted) perf_event_open(..., 0) failed unexpectedly with error 1 (Operation not permitted) Error: You may not have permission to collect system-wide stats. Consider tweaking /proc/sys/kernel/perf_event_paranoid, which controls use of the performance events system by unprivileged users (without CAP_SYS_ADMIN). The current value is 2: -1: Allow use of (almost) all events by all users >= 0: Disallow raw tracepoint access by users without CAP_IOC_LOCK >= 1: Disallow CPU event access by users without CAP_SYS_ADMIN >= 2: Disallow kernel profiling by users without CAP_SYS_ADMIN Needs capabilities to run from a container: Helpful message Although, after setting perf_event_paranoid to -1, it prints the same error...

58. perf & Container Debugging host# strace -fp 26450 [...] [pid 27426] perf_event_open(0x2bfe498, -1, 0, -1, 0) = -1 EPERM (Operation not permitted) [pid 27426] perf_event_open(0x2bfe498, -1, 0, -1, 0) = -1 EPERM (Operation not permitted) [pid 27426] perf_event_open(0x2bfc1a8, -1, 0, -1, PERF_FLAG_FD_CLOEXEC) = -1 EPERM (Operation not permitted) Debugging using strace from the host (as ptrace() is also blocked): bash PID, from which I then ran perf Many different ways to debug this. https://docs.docker.com/engine/security/seccomp/#significant-syscalls-blocked-by-the-default-profile: … …

59. perf, cont. •  Can enable perf_event_open() with: docker run --cap-add sys_admin •  Also need (for kernel symbols): echo 0 > /proc/sys/kernel/kptr_restrict •  perf then "works", and you can make flame graphs. But it sees all CPUs!? •  perf needs to be "container aware", and only see the container's tasks. patch pending: https://lkml.org/lkml/2017/1/12/308 •  Currently easier to run perf from the host (or secure "monitoring" container) •  Via a secure monitoring agent, e.g. Netflix Vector -> CPU Flame Graph •  See earlier slides for steps

60. Advanced Analysis … a few more examples (iosnoop, zfsslower, and btrfsdist shown earlier) 5. Tracing

61. Built-in Linux Tracers Srace (2008+) perf_events (2009+) eBPF (2014+) Some front-ends: •  ftrace: https://github.com/brendangregg/perf-tools •  perf_events: used for CPU flame graphs •  eBPF (aka BPF): https://github.com/iovisor/bcc (Linux 4.4+)

62. ftrace: Overlay FS Function Calls # funccount '*ovl*' Tracing "*ovl*"... Ctrl-C to end. ^C FUNC COUNT ovl_cache_free 3 ovl_xattr_get 3 [...] ovl_fill_merge 339 ovl_path_real 617 ovl_path_upper 777 ovl_update_time 777 ovl_permission 1408 ovl_d_real 1434 ovl_override_creds 1804 Ending tracing... Using ftrace via my perf-tools to count function calls in-kernel context: Each can be a target for further study with kprobes

63. ftrace: Overlay FS Function Tracing # kprobe -s 'p:ovl_fill_merge ctx=%di name=+0(%si):string' Tracing kprobe ovl_fill_merge. Ctrl-C to end. bash-16633 [000] d... 14390771.218973: ovl_fill_merge: (ovl_fill_merge+0x0/0x1f0 [overlay]) ctx=0xffffc90042477db0 name="iostat" bash-16633 [000] d... 14390771.218981: <stack trace> => ovl_fill_merge => ext4_readdir => iterate_dir => ovl_dir_read_merged => ovl_iterate => iterate_dir => SyS_getdents => do_syscall_64 => return_from_SYSCALL_64 […] Using kprobe (perf-tools) to trace ovl_fill_merg() args and stack trace Good for debugging, although dumping all events can cost too much overhead. ftrace has some solutions to this, BPF has more…

64. BPF bytecode Observability Program Kernel tracepoints kprobes uprobes BPF maps per-event data sta=s=cs veriﬁer output sta=c tracing dynamic tracing async copy perf_events sampling, PMCs BPF program event conﬁg aach load Enhanced BPF Tracing Internals

65. BPF: Scheduler Latency 1 host# runqlat -p 20228 10 1 Tracing run queue latency... Hit Ctrl-C to end. usecs : count distribution 0 -> 1 : 0 | | 2 -> 3 : 4 | | 4 -> 7 : 368 |****************************************| 8 -> 15 : 151 |**************** | 16 -> 31 : 22 |** | 32 -> 63 : 14 |* | 64 -> 127 : 19 |** | 128 -> 255 : 0 | | 256 -> 511 : 2 | | 512 -> 1023 : 1 | | This is an app in a Docker container on a system with idle CPU Tracing scheduler events can be costly (high rate), but this BPF program reduces cost by using in-kernel maps to summarize data, and only emits the "count" column to user space.

66. BPF: Scheduler Latency 2 host# runqlat -p 20228 10 1 Tracing run queue latency... Hit Ctrl-C to end. usecs : count distribution 0 -> 1 : 0 | | 2 -> 3 : 0 | | 4 -> 7 : 7 |** | 8 -> 15 : 14 |***** | 16 -> 31 : 0 | | 32 -> 63 : 0 | | 64 -> 127 : 0 | | 128 -> 255 : 0 | | 256 -> 511 : 0 | | 512 -> 1023 : 0 | | 1024 -> 2047 : 0 | | 2048 -> 4095 : 5 |** | 4096 -> 8191 : 6 |** | 8192 -> 16383 : 28 |*********** | 16384 -> 32767 : 59 |*********************** | 32768 -> 65535 : 99 |****************************************| 65536 -> 131071 : 6 |** | 131072 -> 262143 : 2 | | 262144 -> 524287 : 1 | | Now other tenants are using more CPU, and this PID is throttled via CPU shares 8 - 65ms delays

67. BPF: Scheduler Latency 3 host# runqlat --pidnss -m Tracing run queue latency... Hit Ctrl-C to end. ^C pidns = 4026532870 msecs : count distribution 0 -> 1 : 264 |****************************************| 2 -> 3 : 0 | | 4 -> 7 : 0 | | 8 -> 15 : 0 | | 16 -> 31 : 0 | | 32 -> 63 : 0 | | 64 -> 127 : 2 | | […] pidns = 4026532382 msecs : count distribution 0 -> 1 : 646 |****************************************| 2 -> 3 : 18 |* | 4 -> 7 : 48 |** | 8 -> 15 : 17 |* | 16 -> 31 : 150 |********* | 32 -> 63 : 134 |******** | Per-PID namespace histograms (I added this yesterday)

68. BPF: Namespace-ing Tools Walking from the task_struct to the PID namespace ID: task_struct->nsproxy->pid_ns_for_children->ns.inum •  This is unstable, and could break between kernel versions. If it becomes a problem, we'll add a bpf_get_current_pidns() •  Does needs a *task, or bpf_get_current_task() (added in 4.8) •  Can also pull out cgroups, but gets tricker…

69. bcc (BPF) Perf Tools

70. Docker Analysis & Debugging If needed, dockerd can also be analyzed using: •  go execution tracer •  GODEBUG with gctrace and schedtrace •  gdb and Go runtime support •  perf profiling •  bcc/BPF and uprobes Each has pros/cons. bcc/BPF can trace user & kernel events.

71. BPF: dockerd Go Function Counting # funccount '/usr/bin/dockerd:*docker*get*' Tracing 463 functions for "/usr/bin/dockerd:*docker*get*"... Hit Ctrl-C to end. ^C FUNC COUNT github.com/docker/docker/daemon.(*statsCollector).getSystemCPUUsage 3 github.com/docker/docker/daemon.(*Daemon).getNetworkSandboxID 3 github.com/docker/docker/daemon.(*Daemon).getNetworkStats 3 github.com/docker/docker/daemon.(*statsCollector).getSystemCPUUsage.func1 3 github.com/docker/docker/pkg/ioutils.getBuffer 6 github.com/docker/docker/vendor/golang.org/x/net/trace.getBucket 9 github.com/docker/docker/vendor/golang.org/x/net/trace.getFamily 9 github.com/docker/docker/vendor/google.golang.org/grpc.(*ClientConn).getTransport 10 github.com/docker/docker/vendor/github.com/golang/protobuf/proto.getbase 20 github.com/docker/docker/vendor/google.golang.org/grpc/transport.(*http2Client).getStream 30 Detaching... # objdump -tTj .text /usr/bin/dockerd | wc -l 35859 Counting dockerd Go calls in-kernel using BPF that match "*docker*get": 35,859 functions can be traced! Uses uprobes, and needs newer kernels. Warning: will cost overhead at high function rates.

72. BPF: dockerd Go Stack Tracing # stackcount 'p:/usr/bin/dockerd:*/ioutils.getBuffer' Tracing 1 functions for "p:/usr/bin/dockerd:*/ioutils.getBuffer"... Hit Ctrl-C to end. ^C github.com/docker/docker/pkg/ioutils.getBuffer github.com/docker/docker/pkg/broadcaster.(*Unbuffered).Write bufio.(*Reader).writeBuf bufio.(*Reader).WriteTo io.copyBuffer io.Copy github.com/docker/docker/pkg/pools.Copy github.com/docker/docker/container/stream.(*Config).CopyToPipe.func1.1 runtime.goexit dockerd [18176] 110 Detaching... Counting stack traces that led to this ioutils.getBuffer() call: means this stack was seen 110 times Can also trace function arguments, and latency (with some work) http://www.brendangregg.com/blog/2017-01-31/golang-bcc-bpf-function-tracing.html

73. Identify bottlenecks: 1.  In the host vs container, using system metrics 2.  In application code on containers, using CPU flame graphs 3.  Deeper in the kernel, using tracing tools Summary

74. •  http://techblog.netflix.com/2017/04/the-evolution-of-container-usage-at.html •  http://techblog.netflix.com/2016/07/distributed-resource-scheduling-with.html •  https://www.slideshare.net/aspyker/netflix-and-containers-titus •  https://docs.docker.com/engine/admin/runmetrics/#tips-for-high-performance-metric-collection •  https://blog.docker.com/2013/10/gathering-lxc-docker-containers-metrics/ •  https://www.slideshare.net/jpetazzo/anatomy-of-a-container-namespaces-cgroups-some-filesystem-magic-linuxcon •  https://www.youtube.com/watch?v=sK5i-N34im8 Cgroups, namespaces, and beyond •  https://jvns.ca/blog/2016/10/10/what-even-is-a-container/ •  https://blog.jessfraz.com/post/containers-zones-jails-vms/ •  http://blog.alicegoldfuss.com/making-flamegraphs-with-containerized-java/ •  http://www.brendangregg.com/USEmethod/use-linux.html full USE method list •  http://www.brendangregg.com/blog/2017-01-31/golang-bcc-bpf-function-tracing.html •  http://techblog.netflix.com/2015/11/linux-performance-analysis-in-60s.html •  http://queue.acm.org/detail.cfm?id=1809426 latency heat maps •  https://github.com/brendangregg/perf-tools ftrace tools, https://github.com/iovisor/bcc BPF tools References

75. Thank You! http://techblog.netflix.com http://slideshare.net/brendangregg http://www.brendangregg.com bgregg@netflix.com @brendangregg Titus team: @aspyker @anwleung @fabiokung @tomaszbak1974 @amit_joshee @sargun @corindwyer … #dockercon

Container Performance Analysis Brendan Gregg, Netflix

Docker, Inc.

Container Performance Analysis Brendan Gregg, Netflix