Re: Hangs on network hickup

[Jeff Trawick <trawick@xxxxxxxxx>]

2010/6/2 Igor Galić <i.galic@xxxxxxxxxxxxxx>

>

> >root@atvt1uajas001:~# truss -f -p 26794
> >26794:  fcntl(54, F_SETLKW64, 0xFF1E17C0) (sleeping...)
>
> This is consistent with, but not sufficient to be certain of, the
> problem I alluded to.  You'd need to look at the truss of child
> processes not blocked in this call to be reasonably certain.

I only trussed a couple of children, and they all (three or so) showed
the above pattern. The parent was, of course, in a wait() loop.

of the n children which can serve requests, n-1 would have the truss you show (nothing at all wrong) and the remaining one (the one that has obtained the AcceptMutex on which the others await) has the interesting backtrace and truss