From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 9598 invoked by alias); 19 May 2009 19:51:22 -0000 Received: (qmail 9592 invoked by alias); 19 May 2009 19:51:21 -0000 X-SWARE-Spam-Status: No, hits=-2.0 required=5.0 tests=AWL,BAYES_00,SPF_HELO_PASS X-Spam-Status: No, hits=-2.0 required=5.0 tests=AWL,BAYES_00,SPF_HELO_PASS X-Spam-Check-By: sourceware.org X-Spam-Checker-Version: SpamAssassin 3.2.5 (2008-06-10) on bastion2.fedora.phx.redhat.com Subject: cluster: RHEL5 - rgmanager: Allow reboot if main proc. is killed To: cluster-cvs-relay@redhat.com X-Project: Cluster Project X-Git-Module: cluster.git X-Git-Refname: refs/heads/RHEL5 X-Git-Reftype: branch X-Git-Oldrev: df24d6192e1a60d91c507c7dc5e8de1a4eaa55a4 X-Git-Newrev: d727e39ecd607b879ec3dc8841d599131ee7638d From: Lon Hohberger Message-Id: <20090519195053.607BB120152@lists.fedorahosted.org> Date: Tue, 19 May 2009 19:51:00 -0000 X-Scanned-By: MIMEDefang 2.58 on 172.16.52.254 Mailing-List: contact cluster-cvs-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Post: List-Help: , Sender: cluster-cvs-owner@sourceware.org X-SW-Source: 2009-q2/txt/msg00334.txt.bz2 Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=d727e39ecd607b879ec3dc8841d599131ee7638d Commit: d727e39ecd607b879ec3dc8841d599131ee7638d Parent: df24d6192e1a60d91c507c7dc5e8de1a4eaa55a4 Author: Lon Hohberger AuthorDate: Tue May 19 15:45:13 2009 -0400 Committer: Lon Hohberger CommitterDate: Tue May 19 15:45:13 2009 -0400 rgmanager: Allow reboot if main proc. is killed The Linux OOM killer uses SIGKILL to destroy processes. While rgmanager isn't likely to die due to high memory pressure due to a low 'badness' score, inadvertently dying and not rebooting the node can have unintended consequences. Resolves: 488072 Signed-off-by: Lon Hohberger --- rgmanager/src/daemons/watchdog.c | 24 ++++++++++++++---------- 1 files changed, 14 insertions(+), 10 deletions(-) diff --git a/rgmanager/src/daemons/watchdog.c b/rgmanager/src/daemons/watchdog.c index c748382..9e3e9ac 100644 --- a/rgmanager/src/daemons/watchdog.c +++ b/rgmanager/src/daemons/watchdog.c @@ -21,6 +21,7 @@ #include #include #include +#include #include #include @@ -68,6 +69,7 @@ watchdog_init(void) return parent; redirect_signals(); + mlockall(MCL_CURRENT); /* shouldn't need MCL_FUTURE */ while (1) { if (waitpid(child, &status, 0) <= 0) @@ -78,20 +80,22 @@ watchdog_init(void) if (WIFSIGNALED(status)) { if (WTERMSIG(status) == SIGKILL) { - clulog(LOG_CRIT, "Watchdog: Daemon killed, exiting\n"); - raise(SIGKILL); - while(1) ; + /* Assume the admin did a 'killall' - it will + * kill us within a couple of seconds. If + * we are still alive after this sleep, it + * could have been the OOM killer killing + * rgmanager proper and we need to reboot. + */ + sleep(3); } - else { #ifdef DEBUG - clulog(LOG_CRIT, "Watchdog: Daemon died, but not rebooting because DEBUG is set\n"); + clulog(LOG_CRIT, "Watchdog: Daemon died, but not rebooting because DEBUG is set\n"); #else - clulog(LOG_CRIT, "Watchdog: Daemon died, rebooting...\n"); - sync(); - reboot(RB_AUTOBOOT); + clulog(LOG_CRIT, "Watchdog: Daemon died, rebooting...\n"); + sync(); + reboot(RB_AUTOBOOT); #endif - exit(255); - } + exit(255); } } }