You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Refer to #3753
BackGround:
I discovered this problem by an unusual path - I have a multi-threaded search, and because of the danger
of thread deadlock bugs, I monitor the searching threads in another thread. and note if they seem to stop
making progress. I can only do that indirectly, by monitoring their explicitly published progress metrics.
... so while the actual problem was that the garbage collector had crashed and processes were piling up
in "waiting for gc" state, I reported it as a "thread deadlock". The underlying problem, in other circumstances,
would be manifest as random lockup/freeze, or worse - very hard to debug.
Request:
provide more information about the state of threads. At least "runnable" "waiting for synchronization monitor" "dead"
and similar states ought to be possible. Anything else would be gravy, ideally complete information about which
threads are waiting for monitors held by which other threads and stack traces. Then a top level watchdog thread
could be run to provide an overall gods-eye view of the application.
The text was updated successfully, but these errors were encountered:
Refer to #3753
BackGround:
I discovered this problem by an unusual path - I have a multi-threaded search, and because of the danger
of thread deadlock bugs, I monitor the searching threads in another thread. and note if they seem to stop
making progress. I can only do that indirectly, by monitoring their explicitly published progress metrics.
... so while the actual problem was that the garbage collector had crashed and processes were piling up
in "waiting for gc" state, I reported it as a "thread deadlock". The underlying problem, in other circumstances,
would be manifest as random lockup/freeze, or worse - very hard to debug.
Request:
provide more information about the state of threads. At least "runnable" "waiting for synchronization monitor" "dead"
and similar states ought to be possible. Anything else would be gravy, ideally complete information about which
threads are waiting for monitors held by which other threads and stack traces. Then a top level watchdog thread
could be run to provide an overall gods-eye view of the application.
The text was updated successfully, but these errors were encountered: