Hi
I have a problem with my backend mail server. Sometimes it stops and
does nothing
for few or more seconds ( sometimes minutes ).
Using dtrace I found that server stops on holdlwps during socket close.
I suspect that in the same time
this socket is used by another syscall ( shutdown ) in other thread -
that is why holdlwps is called.
The backend server has about 200 threads and is used to tranfer data
from NFS storage to frontends.
Server is T2000 ( 32 cores ).
Why does holdlwps takes so much time ?
Regards
Lukas Karwacki
Dtrace log ( time in ms ):
hold: 80531 [80529 1]
hold: 73873 [73871 1]
hold: 74990 [74988 1]
Dtrace script:
fbt::holdlwps:entry
/execname=="wpmailio_1"/
{
self->time = timestamp;
self->tak = 1;
}
fbt::holdlwps:return
/self->tak == 1/
{
self->t1 = timestamp;
}
fbt::continuelwps:entry
/self->tak == 1/
{
self->t2 = timestamp;
}
fbt::continuelwps:return
/self->tak == 1/
{
printf("hold: %d [%d %d]\n",(timestamp - self->time)/1000000,
(self->t1 - self->time)/1000000,
(timestamp - self->t2)/1000000);
self->tak = 0;
}
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
