Feature #17853
closedAdd Thread#thread_id
Description
Abstract¶
New method Thread#thread_id
to get associated native thread id (LWP.) It might return nil
if OS doesn't support thread id or equivalent.
Background¶
When I tried to investigate which Ruby thread of an application is busy, I did the following steps
- checked the CPU usage of the Ruby application's threads using
ps -eLf
ortop
(with H key) and got which thread is busy - dumped all the threads of the application using https://github.com/frsyuki/sigdump
- tried to find a busy thread in the thread dump result, but the thread dump doesn't contain thread id...
Thread
class itself has no method to get associated thread id. If the class has #thread_id
or something, I can create a PR on sigdump
project to add thread id
in thread dump output to make investigations with thread dump much easier.
Thread#name
may seem an alternative to Thread#thread_id
, but Thread#name
just returns a value that's set through Thread#name=
while Thread#thread_id
returns a thread id that OS or something has assigned.
In case of Java, thread dump created by jcmd ${pid} Thread.print
contains nid
field which is an associated native thread id
"http-bio-80-exec-77" daemon prio=6 tid=0x0000000026f29000 nid=0xbd0 runnable [0x0000000020c7f000]
java.lang.Thread.State: RUNNABLE
at java.net.SocketInputStream.socketRead0(Native Method)
at java.net.SocketInputStream.read(Unknown Source)
at java.net.SocketInputStream.read(Unknown Source)
at org.apache.coyote.http11.InternalInputBuffer.fill(InternalInputBuffer.java:516)
at org.apache.coyote.http11.InternalInputBuffer.fill(InternalInputBuffer.java:501)
at org.apache.coyote.http11.Http11Processor.setRequestLineReadTimeout(Http11Processor.java:167)
at org.apache.coyote.http11.AbstractHttp11Processor.process(AbstractHttp11Processor.java:946)
at org.apache.coyote.AbstractProtocol$AbstractConnectionHandler.process(AbstractProtocol.java:607)
at org.apache.tomcat.util.net.JIoEndpoint$SocketProcessor.run(JIoEndpoint.java:315)
- locked <0x00000007b16e3e88> (a org.apache.tomcat.util.net.SocketWrapper)
at java.util.concurrent.ThreadPoolExecutor.runWorker(Unknown Source)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(Unknown Source)
at java.lang.Thread.run(Unknown Source)
Updated by naruse (Yui NARUSE) over 3 years ago
- Related to Feature #11251: Thread#name and Thread#name= added
Updated by komamitsu (Mitsunori Komatsu) over 3 years ago
- Description updated (diff)
Updated by komamitsu (Mitsunori Komatsu) over 3 years ago
- Description updated (diff)
Updated by yui-knk (Kaneko Yuichiro) over 3 years ago
I investigated Python for reference. Current CPython supports threading.get_native_id
(https://docs.python.org/3/library/threading.html#threading.get_native_id) which returns native Thread ID.
Implementation details are
- https://github.com/python/cpython/blob/bb3e0c240bc60fe08d332ff5955d54197f79751c/Python/thread_pthread.h#L332-L356
- https://github.com/python/cpython/blob/bb3e0c240bc60fe08d332ff5955d54197f79751c/Python/thread_nt.h#L241-L252
This method can be used as below
# cat thread.py
import threading
import time
import os
def p():
while True:
print(threading.get_native_id())
time.sleep(1)
print(os.getpid())
thread1 = threading.Thread(target=p)
thread2 = threading.Thread(target=p)
thread1.start()
thread2.start()
thread1.join()
thread2.join()
# python thread.py
15 # <- pid
16 # <- Thread id
17 # <- Thread id
# ps -efL
UID PID PPID LWP C NLWP STIME TTY TIME CMD
root 1 0 1 0 1 07:28 pts/0 00:00:00 /bin/bash
root 9 0 9 0 1 07:29 pts/1 00:00:00 /bin/bash
root 15 1 15 0 3 07:30 pts/0 00:00:00 python thread.py
root 15 1 16 0 3 07:30 pts/0 00:00:00 python thread.py
root 15 1 17 0 3 07:30 pts/0 00:00:00 python thread.py
root 19 9 19 0 1 07:30 pts/1 00:00:00 ps -efL
Updated by ivoanjo (Ivo Anjo) over 3 years ago
+1 This would be useful for Datadog's ddtrace gem as well: currently we need to do quite a bit of hackery and monkey-patching to get the per-thread cpu clock information, and if we had the native id, we could use that instead to call Process.clock_gettime
.
Updated by ko1 (Koichi Sasada) over 3 years ago
Two concerns.
(1) name should include native
, os
or something to represent it is platform value. Maybe native_id
same as Python's API is enough. Ruby can set internal ID for threads.
(2) In future the thread model can be changed (1:1 -> M:N). In this case, native_id
should be changed (or nil
if the ruby thread is not running on native thread).
Updated by komamitsu (Mitsunori Komatsu) over 3 years ago
native_id
sounds better to me. I agree with it.
As for m:n thread model
, in my opinion, either volatile value or nil
(this can happen when it's waiting an event in select/epoll/kqueue
without assigning any thread?) sounds good to me since what I want to do is just mapping the result of Ruby Thread information to native thread resource. current_native_id
might be better since it sounds like the attribute/property can be easily changed.
Updated by naruse (Yui NARUSE) over 3 years ago
Created a proposed implementation:
https://github.com/ruby/ruby/compare/master...nurse:native_thread_id
Updated by naruse (Yui NARUSE) over 3 years ago
- Status changed from Open to Closed
Applied in changeset git|46655156dcc37509dcb69fcd0717c110eb1c624a.
Add Thread#native_thread_id [Feature #17853]
Updated by ivoanjo (Ivo Anjo) over 3 years ago
Thanks a lot for building this, @naruse (Yui NARUSE)!
If I can ask for a tiny addition, it would be really cool to get the native_thread_id
, if available, in Thread#inspect
:)
Updated by nobu (Nobuyoshi Nakada) over 3 years ago
diff --git i/thread.c w/thread.c
index cbef44a9d4a..c627e5c8f2a 100644
--- i/thread.c
+++ w/thread.c
@@ -3456,6 +3456,14 @@ rb_thread_to_s(VALUE thread)
if (!NIL_P(target_th->name)) {
rb_str_catf(str, "@%"PRIsVALUE, target_th->name);
}
+#if USE_NATIVE_THREAD_NATIVE_THREAD_ID
+ if (!rb_threadptr_dead(target_th)) {
+ VALUE native_id = native_thread_native_thread_id(target_th);
+ if (!NIL_P(native_id)) {
+ rb_str_catf(str, "(%"PRIsVALUE")", native_id);
+ }
+ }
+#endif
if ((loc = threadptr_invoke_proc_location(target_th)) != Qnil) {
rb_str_catf(str, " %"PRIsVALUE":%"PRIsVALUE,
RARRAY_AREF(loc, 0), RARRAY_AREF(loc, 1));
Updated by ivoanjo (Ivo Anjo) over 3 years ago
@nobu (Nobuyoshi Nakada) that looks great!
Updated by komamitsu (Mitsunori Komatsu) over 3 years ago
Thank you, @naruse (Yui NARUSE) and @nobu (Nobuyoshi Nakada)!
Updated by Eregon (Benoit Daloze) over 1 year ago
FWIW, it seems the Thread#inspect
change wasn't merged.