Some more test I have do:
1. running qemu with TCMALLOC_MAX_TOTAL_THREAD_CACHE_BYTES 256MB, still got
problem
2. prevent cpu go into C3 and C6 state, still got problem
3. running qemu with aio=native, still got problem
--
You received this bug notification because you are a member of qemu-
devel
Here are 'perf top -p `pgrep qemu` -a` output; I met tcmalloc problem on the
osd host and fix it with a larger TCMALLOC_MAX_TOTAL_THREAD_CACHE_BYTES. The
perf top of tcmalloc problem is a little bit different of my problem.
--
Sa
Here are gdb oupout with librbd1-dbg and librados2-dbg.
-
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
0x7ff8cf8dddff in ppoll () from /lib/x86_64-linux-gnu/libc.
Public bug reported:
I got an IO problem. When running Qemu + ceph(use librbd), and do a random IO
benchmark or some high load random IO test, it will exhaust all my host cpu on
%sy cpu.
It doesn’t happen all the time, but when it appear it will reproduce every time
I start a random IO benchmar
