Bacula-users

[Bacula-users] should SIGTHAW be ignored, rather than cause termination?

2008-04-16 12:12:41
Subject: [Bacula-users] should SIGTHAW be ignored, rather than cause termination?
From: John Jorgensen <jorgnsn AT lcd.uregina DOT ca>
To: bacula-users AT lists.sourceforge DOT net
Date: Wed, 16 Apr 2008 09:43:46 -0600
This morning one of our Bacula clients, running bacula-fd 2.2.8
on Solaris 5.9, was unreachable.  It turned that somebody had
suspended and resumed the machine:

  Apr 15 16:27:40 name-changed cpr: [ID 535284 kern.notice] System is being 
suspended
  Apr 15 16:28:39 name-changed cpr: [ID 583038 kern.notice] System has been 
resumed.
  Apr 15 16:28:43 name-changed hme: [ID 517527 kern.info] SUNW,hme0 : Internal 
Transceiver Selected.
  Apr 15 16:28:43 name-changed hme: [ID 517527 kern.info] SUNW,hme0 :   100 
Mbps Full-Duplex Link Up
  Apr 15 16:28:46 name-changed bacula-fd[13873]: [ID 702911 daemon.error] 
name-changed-fd: Fatal Error because: Bacula interrupted by signal 35: 
Checkpoint Thaw

The Solaris cpr(7) man page says that when the system is
"resume"d, the processes whose state is restored are sent SIGTHAW
to get the chance to do whatever housekeeping they might require
to resume operations, but that most can simply ignore the signal.

So at first I thought that SIGTHAW might be a Solaris-ism which
bacula does not know about, and that it was bacula's default signal
handler which was exiting, rather than just continuing
execution.

But src/lib/signal.c does explicitly set the SIGTHAW handler to
sighandle (which I gather exits) rather than sigignore.

Is this an example of the bacula developers simply making
conservative assumptions, or do we know that bacula-fd could not
successfully continue execution after resumption?

-- 
John Jorgensen  LCD System Administrator  jorgnsn AT lcd.uregina DOT ca
                                          306.337.2344


-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users

<Prev in Thread] Current Thread [Next in Thread>
  • [Bacula-users] should SIGTHAW be ignored, rather than cause termination?, John Jorgensen <=