strace problem (Was: Not to use SCAN in 11gR2 RAC)
From: Yong Huang <yong321_at_yahoo.com>
Date: Thu, 22 Oct 2009 13:02:08 -0700 (PDT)
Message-ID: <916147.70278.qm_at_web80603.mail.mud.yahoo.com>
> One interesting observation is that, if you run strace on either
> the SCAN or regular listener, as soon as you Control-C to stop
> strace, the listener hangs when you tnsping or sqlplus to it.
> (Just kill it and it'll be respawned immediately.)
Date: Thu, 22 Oct 2009 13:02:08 -0700 (PDT)
Message-ID: <916147.70278.qm_at_web80603.mail.mud.yahoo.com>
> One interesting observation is that, if you run strace on either
> the SCAN or regular listener, as soon as you Control-C to stop
> strace, the listener hangs when you tnsping or sqlplus to it.
> (Just kill it and it'll be respawned immediately.)
Correction. It's a problem of strace. Details are at https://www.redhat.com/archives/redhat-list/2009-October/msg00122.html
The behavior may depend on the combination of strace version and Linux kernel version. In one case, Control-C when you run strace won't exit strace. You have to kill it, and the traced process continues to run. In another case (as in my last message above), Control-C exits strace but will send a STOP signal to the traced process, causing it to hang. The least disruptive way to resume its execution is to send a CONTinue signal to it.
Manpage for strace actually warns about this problem (last part of NOTE). The "cousin" of strace, ltrace, doesn't have this problem.
Yong Huang
-- http://www.freelists.org/webpage/oracle-lReceived on Thu Oct 22 2009 - 15:02:08 CDT