[kudu-CR] KUDU-2215. kernel stack watchdog: avoid blocking thread exit

[Todd Lipcon (Code Review)]

Todd Lipcon has submitted this change and it was merged. ( 
http://gerrit.cloudera.org:8080/8536 )

Change subject: KUDU-2215. kernel_stack_watchdog: avoid blocking thread exit
..

KUDU-2215. kernel_stack_watchdog: avoid blocking thread exit

This changes the stack watchdog so that thread unregistration no longer
blocks if the watchdog thread is in the middle of dumping a stack.

This is to try to avoid cases where a user thread is waiting to join on
another thread, but that thread is blocked due to watchdog interference.

A new stress-test/benchmark verifies the improvement. It simulates slow
stack trace collection by injecting latency into the watchdog thread,
and then starts and joins threads in a loop for 5 seconds. Without the
fix, it was only able to start about 1000 threads/second, whereas with
the fix it's able to start 10,000 threads/second.

Change-Id: Ib6a349666e8484c00b2f43c5918205ec1a4c09ab
Reviewed-on: http://gerrit.cloudera.org:8080/8536
Tested-by: Kudu Jenkins
Reviewed-by: Andrew Wong 
---
M src/kudu/util/fault_injection.cc
M src/kudu/util/fault_injection.h
M src/kudu/util/kernel_stack_watchdog.cc
M src/kudu/util/kernel_stack_watchdog.h
M src/kudu/util/stack_watchdog-test.cc
5 files changed, 149 insertions(+), 21 deletions(-)

Approvals:
  Kudu Jenkins: Verified
  Andrew Wong: Looks good to me, approved

--
To view, visit http://gerrit.cloudera.org:8080/8536
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: merged
Gerrit-Change-Id: Ib6a349666e8484c00b2f43c5918205ec1a4c09ab
Gerrit-Change-Number: 8536
Gerrit-PatchSet: 5
Gerrit-Owner: Todd Lipcon 
Gerrit-Reviewer: Andrew Wong 
Gerrit-Reviewer: Andrew Wong 
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Tidy Bot
Gerrit-Reviewer: Todd Lipcon 

[kudu-CR] KUDU-2215. kernel stack watchdog: avoid blocking thread exit

[Andrew Wong (Code Review)]

Andrew Wong has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/8536 )

Change subject: KUDU-2215. kernel_stack_watchdog: avoid blocking thread exit
..


Patch Set 4: Code-Review+2


--
To view, visit http://gerrit.cloudera.org:8080/8536
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Ib6a349666e8484c00b2f43c5918205ec1a4c09ab
Gerrit-Change-Number: 8536
Gerrit-PatchSet: 4
Gerrit-Owner: Todd Lipcon 
Gerrit-Reviewer: Andrew Wong 
Gerrit-Reviewer: Andrew Wong 
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Tidy Bot
Gerrit-Reviewer: Todd Lipcon 
Gerrit-Comment-Date: Sat, 18 Nov 2017 04:27:00 +
Gerrit-HasComments: No

[kudu-CR] KUDU-2215. kernel stack watchdog: avoid blocking thread exit

[Todd Lipcon (Code Review)]

Hello Tidy Bot, Andrew Wong, Kudu Jenkins, Andrew Wong,

I'd like you to reexamine a change. Please visit

http://gerrit.cloudera.org:8080/8536

to look at the new patch set (#4).

Change subject: KUDU-2215. kernel_stack_watchdog: avoid blocking thread exit
..

KUDU-2215. kernel_stack_watchdog: avoid blocking thread exit

This changes the stack watchdog so that thread unregistration no longer
blocks if the watchdog thread is in the middle of dumping a stack.

This is to try to avoid cases where a user thread is waiting to join on
another thread, but that thread is blocked due to watchdog interference.

A new stress-test/benchmark verifies the improvement. It simulates slow
stack trace collection by injecting latency into the watchdog thread,
and then starts and joins threads in a loop for 5 seconds. Without the
fix, it was only able to start about 1000 threads/second, whereas with
the fix it's able to start 10,000 threads/second.

Change-Id: Ib6a349666e8484c00b2f43c5918205ec1a4c09ab
---
M src/kudu/util/fault_injection.cc
M src/kudu/util/fault_injection.h
M src/kudu/util/kernel_stack_watchdog.cc
M src/kudu/util/kernel_stack_watchdog.h
M src/kudu/util/stack_watchdog-test.cc
5 files changed, 149 insertions(+), 21 deletions(-)


  git pull ssh://gerrit.cloudera.org:29418/kudu refs/changes/36/8536/4
--
To view, visit http://gerrit.cloudera.org:8080/8536
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: newpatchset
Gerrit-Change-Id: Ib6a349666e8484c00b2f43c5918205ec1a4c09ab
Gerrit-Change-Number: 8536
Gerrit-PatchSet: 4
Gerrit-Owner: Todd Lipcon 
Gerrit-Reviewer: Andrew Wong 
Gerrit-Reviewer: Andrew Wong 
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Tidy Bot
Gerrit-Reviewer: Todd Lipcon 

[kudu-CR] KUDU-2215. kernel stack watchdog: avoid blocking thread exit

[Todd Lipcon (Code Review)]

Hello Tidy Bot, Andrew Wong, Kudu Jenkins, Andrew Wong,

I'd like you to reexamine a change. Please visit

http://gerrit.cloudera.org:8080/8536

to look at the new patch set (#3).

Change subject: KUDU-2215. kernel_stack_watchdog: avoid blocking thread exit
..

KUDU-2215. kernel_stack_watchdog: avoid blocking thread exit

This changes the stack watchdog so that thread unregistration no longer
blocks if the watchdog thread is in the middle of dumping a stack.

This is to try to avoid cases where a user thread is waiting to join on
another thread, but that thread is blocked due to watchdog interference.

A new stress-test/benchmark verifies the improvement. It simulates slow
stack trace collection by injecting latency into the watchdog thread,
and then starts and joins threads in a loop for 5 seconds. Without the
fix, it was only able to start about 1000 threads/second, whereas with
the fix it's able to start 10,000 threads/second.

Change-Id: Ib6a349666e8484c00b2f43c5918205ec1a4c09ab
---
M src/kudu/util/fault_injection.cc
M src/kudu/util/fault_injection.h
M src/kudu/util/kernel_stack_watchdog.cc
M src/kudu/util/kernel_stack_watchdog.h
M src/kudu/util/stack_watchdog-test.cc
5 files changed, 148 insertions(+), 21 deletions(-)


  git pull ssh://gerrit.cloudera.org:29418/kudu refs/changes/36/8536/3
--
To view, visit http://gerrit.cloudera.org:8080/8536
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: newpatchset
Gerrit-Change-Id: Ib6a349666e8484c00b2f43c5918205ec1a4c09ab
Gerrit-Change-Number: 8536
Gerrit-PatchSet: 3
Gerrit-Owner: Todd Lipcon 
Gerrit-Reviewer: Andrew Wong 
Gerrit-Reviewer: Andrew Wong 
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Tidy Bot
Gerrit-Reviewer: Todd Lipcon 

[kudu-CR] KUDU-2215. kernel stack watchdog: avoid blocking thread exit

[Andrew Wong (Code Review)]

Andrew Wong has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/8536 )

Change subject: KUDU-2215. kernel_stack_watchdog: avoid blocking thread exit
..


Patch Set 1:

(4 comments)

Injection looks good, mostly nits here

http://gerrit.cloudera.org:8080/#/c/8536/1/src/kudu/util/kernel_stack_watchdog.h
File src/kudu/util/kernel_stack_watchdog.h:

http://gerrit.cloudera.org:8080/#/c/8536/1/src/kudu/util/kernel_stack_watchdog.h@176
PS1, Line 176: static void ThreadExiting(void* tls_void);
nit: maybe doc here that this is used internally by CreateTLS to create 
thread-local destructors?


http://gerrit.cloudera.org:8080/#/c/8536/1/src/kudu/util/kernel_stack_watchdog.cc
File src/kudu/util/kernel_stack_watchdog.cc:

http://gerrit.cloudera.org:8080/#/c/8536/1/src/kudu/util/kernel_stack_watchdog.cc@151
PS1, Line 151:  vector to_delete;
 : {
 :   lock_guard l(tls_lock_);
 :   to_delete.swap(pending_delete_);
 :   tls_map_copy = tls_by_tid_;
 : }
 : to_delete.clear();
Is it important to document somewhere that the pending TLS instances only get 
d'ted in RunThread? Or is that more of an implementation detail.


http://gerrit.cloudera.org:8080/#/c/8536/1/src/kudu/util/stack_watchdog-test.cc
File src/kudu/util/stack_watchdog-test.cc:

http://gerrit.cloudera.org:8080/#/c/8536/1/src/kudu/util/stack_watchdog-test.cc@139
PS1, Line 139: std::t
nit: drop std::


http://gerrit.cloudera.org:8080/#/c/8536/1/src/kudu/util/stack_watchdog-test.cc@139
PS1, Line 139: i
nit: could replace with `started % threads.size()`? One fewer variable to think 
about, as trivial as it is



--
To view, visit http://gerrit.cloudera.org:8080/8536
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Ib6a349666e8484c00b2f43c5918205ec1a4c09ab
Gerrit-Change-Number: 8536
Gerrit-PatchSet: 1
Gerrit-Owner: Todd Lipcon 
Gerrit-Reviewer: Andrew Wong 
Gerrit-Reviewer: Andrew Wong 
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Tidy Bot
Gerrit-Comment-Date: Mon, 13 Nov 2017 23:45:13 +
Gerrit-HasComments: Yes

[kudu-CR] KUDU-2215. kernel stack watchdog: avoid blocking thread exit

[Todd Lipcon (Code Review)]

Hello Andrew Wong,

I'd like you to do a code review. Please visit

http://gerrit.cloudera.org:8080/8536

to review the following change.


Change subject: KUDU-2215. kernel_stack_watchdog: avoid blocking thread exit
..

KUDU-2215. kernel_stack_watchdog: avoid blocking thread exit

This changes the stack watchdog so that thread unregistration no longer
blocks if the watchdog thread is in the middle of dumping a stack.

This is to try to avoid cases where a user thread is waiting to join on
another thread, but that thread is blocked due to watchdog interference.

A new stress-test/benchmark verifies the improvement. It simulates slow
stack trace collection by injecting latency into the watchdog thread,
and then starts and joins threads in a loop for 5 seconds. Without the
fix, it was only able to start about 1000 threads/second, whereas with
the fix it's able to start 10,000 threads/second.

Change-Id: Ib6a349666e8484c00b2f43c5918205ec1a4c09ab
---
M src/kudu/util/fault_injection.cc
M src/kudu/util/fault_injection.h
M src/kudu/util/kernel_stack_watchdog.cc
M src/kudu/util/kernel_stack_watchdog.h
M src/kudu/util/stack_watchdog-test.cc
5 files changed, 139 insertions(+), 17 deletions(-)



  git pull ssh://gerrit.cloudera.org:29418/kudu refs/changes/36/8536/1
--
To view, visit http://gerrit.cloudera.org:8080/8536
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: newchange
Gerrit-Change-Id: Ib6a349666e8484c00b2f43c5918205ec1a4c09ab
Gerrit-Change-Number: 8536
Gerrit-PatchSet: 1
Gerrit-Owner: Todd Lipcon 
Gerrit-Reviewer: Andrew Wong