msg/async: print error log if add_event fail #17102
Conversation
|
@xiexingguo @joscollin I think this is necessary, how do you think of it ? |
src/os/bluestore/KernelDevice.cc
Outdated
| << " with " << (ioc->num_running.load() - 1) | ||
| << " aios left" << dendl; | ||
| assert(r >= 0); | ||
| } | ||
|
|
There was a problem hiding this comment.
This prints 2 times when r < 0 right ? Please see https://github.com/mychoxin/ceph/blob/deb79f35fde7c6c49083685673562e4fe18bb506/src/os/bluestore/KernelDevice.cc#L365
mychoxin
force-pushed
the
yuanxin_print_log_if_add_event_fail
branch
2 times, most recently
from
549d598
to
4ef24b9
Compare
src/os/bluestore/KernelDevice.cc
Outdated
| << " with " << (ioc->num_running.load() - 1) | ||
| << " aios left" << dendl; | ||
| assert(r >= 0); | ||
| } | ||
| dout(10) << __func__ << " finished aio " << aio[i] << " r " << r |
There was a problem hiding this comment.
@joscollin at first, for dout(10) when the error happens this log with 10 level not always prints, because opening low level log impacts the performance, for the second, in order to avoid printing log twice when we exactly opened log of this level, i just handle the error before dout(10), how do you think of it ?
There was a problem hiding this comment.
just change the dout(10) to dout(r < 0 ? -1 : 10). (dout(-1) is equivalent to derr)
mychoxin
force-pushed
the
yuanxin_print_log_if_add_event_fail
branch
from
4ef24b9
to
31cc06c
Compare
src/os/bluestore/KernelDevice.cc
Outdated
| << " with " << (ioc->num_running.load() - 1) | ||
| << " aios left" << dendl; | ||
| assert(r >= 0); | ||
| } | ||
| dout(10) << __func__ << " finished aio " << aio[i] << " r " << r |
There was a problem hiding this comment.
just change the dout(10) to dout(r < 0 ? -1 : 10). (dout(-1) is equivalent to derr)
mychoxin
force-pushed
the
yuanxin_print_log_if_add_event_fail
branch
from
31cc06c
to
16ad044
Compare
|
@liewegas I modified it, please have a look. |
Signed-off-by: mychoxin <mychoxin@gmail.com>
mychoxin
force-pushed
the
yuanxin_print_log_if_add_event_fail
branch
from
16ad044
to
52c4c13
Compare
|
@joscollin @liewegas |
|
Jenkins retest this please |
|
@liewegas please let me know if there is exists any problems. |
I ran into some osd assert fail, I thought it resulted of /proc/sys/fs/aio-max-nr being exhausted, but I can not
reproduce it and verify it (not so easy), so I think here we should add some error log, which makes us know what the error code is.