All of lore.kernel.org
 help / color / mirror / Atom feed
From: lhh@sourceware.org <lhh@sourceware.org>
To: cluster-devel.redhat.com
Subject: [Cluster-devel] cluster/fence/fenced fd.h main.c recover.c
Date: 29 Jan 2007 20:30:26 -0000	[thread overview]
Message-ID: <20070129203026.13144.qmail@sourceware.org> (raw)

CVSROOT:	/cvs/cluster
Module name:	cluster
Branch: 	RHEL5
Changes by:	lhh at sourceware.org	2007-01-29 20:30:25

Modified files:
	fence/fenced   : fd.h main.c recover.c 

Log message:
	Add manual override for fenced to RHEL5 branch; patch is a merge from HEAD branch

Patches:
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/fence/fenced/fd.h.diff?cvsroot=cluster&only_with_tag=RHEL5&r1=1.24.2.1&r2=1.24.2.2
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/fence/fenced/main.c.diff?cvsroot=cluster&only_with_tag=RHEL5&r1=1.38.2.2&r2=1.38.2.3
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/fence/fenced/recover.c.diff?cvsroot=cluster&only_with_tag=RHEL5&r1=1.25.2.1&r2=1.25.2.2

--- cluster/fence/fenced/fd.h	2006/12/01 19:17:46	1.24.2.1
+++ cluster/fence/fenced/fd.h	2007/01/29 20:30:25	1.24.2.2
@@ -53,6 +53,7 @@
 #define DEFAULT_POST_JOIN_DELAY	6
 #define DEFAULT_POST_FAIL_DELAY	0
 #define DEFAULT_CLEAN_START	0
+#define DEFAULT_OVERRIDE_PATH	"/var/run/cluster/fenced_override"
 #define FENCED_SOCK_PATH	"fenced_socket"
 
 extern char			*prog_name;
@@ -131,10 +132,12 @@
 {
 	int post_join_delay;
 	int post_fail_delay;
+	char *override_path;
 	int8_t clean_start;
 	int8_t post_join_delay_opt;
 	int8_t post_fail_delay_opt;
 	int8_t clean_start_opt;
+	int8_t override_path_opt;
 };
 
 #define FDFL_RUN        (0)
--- cluster/fence/fenced/main.c	2006/12/01 15:27:50	1.38.2.2
+++ cluster/fence/fenced/main.c	2007/01/29 20:30:25	1.38.2.3
@@ -15,7 +15,7 @@
 #include "ccs.h"
 #include "copyright.cf"
 
-#define OPTION_STRING			("cj:f:Dn:hVSw")
+#define OPTION_STRING			("cj:f:Dn:O:hVSw")
 #define LOCKFILE_NAME			"/var/run/fenced.pid"
 
 struct client {
@@ -145,6 +145,23 @@
 			free(str);
 	}
 
+	if (comline.override_path_opt == FALSE) {
+		str = NULL;
+		memset(path, 0, 256);
+		sprintf(path, "/cluster/fence_daemon/@override_path");
+
+		error = ccs_get(cd, path, &str);
+		if (!error)
+			/* XXX These are not explicitly freed on exit; if
+			   we decide to make fenced handle SIGHUP at a later
+			   time, we will need to free this. */
+			comline.override_path = strdup(str);
+		else
+			comline.override_path = strdup(DEFAULT_OVERRIDE_PATH);
+		if (str)
+			free(str);
+	}
+
 	log_debug("delay post_join %ds post_fail %ds",
 		  comline.post_join_delay, comline.post_fail_delay);
 
@@ -500,6 +517,8 @@
 				   DEFAULT_POST_JOIN_DELAY);
 	printf("  -f <secs>	Post-fail fencing delay (default %d)\n",
 				   DEFAULT_POST_FAIL_DELAY);
+	printf("  -O <path>    Override path (default %s)\n",
+	       			   DEFAULT_OVERRIDE_PATH);
 	printf("  -D	       Enable debugging code and don't fork\n");
 	printf("  -h	       Print this help, then exit\n");
 	printf("  -V	       Print program version information, then exit\n");
@@ -547,6 +566,8 @@
 	int cont = TRUE;
 	int optchar;
 
+	comline->override_path_opt = FALSE;
+	comline->override_path = NULL;
 	comline->post_join_delay_opt = FALSE;
 	comline->post_fail_delay_opt = FALSE;
 	comline->clean_start_opt = FALSE;
@@ -571,6 +592,11 @@
 			comline->post_fail_delay_opt = TRUE;
 			break;
 
+		case 'O':
+			comline->override_path = strdup(optarg);
+			comline->override_path_opt = TRUE;
+			break;
+
 		case 'D':
 			daemon_debug_opt = TRUE;
 			break;
--- cluster/fence/fenced/recover.c	2006/12/01 19:17:46	1.25.2.1
+++ cluster/fence/fenced/recover.c	2007/01/29 20:30:25	1.25.2.2
@@ -13,6 +13,9 @@
 
 #include "fd.h"
 #include "ccs.h"
+#include <sys/time.h>
+#include <sys/types.h>
+#include <sys/select.h>
 
 extern int our_nodeid;
 extern commandline_t comline;
@@ -212,6 +215,79 @@
 	return num_victims;
 }
 
+static inline void close_override(int *fd, char *path)
+{
+	unlink(path);
+	if (fd && *fd >= 0)
+		close(*fd);
+	*fd = -1;
+}
+
+static int open_override(char *path)
+{
+	int ret;
+	mode_t om;
+
+	om = umask(077);
+	ret = mkfifo(path, (S_IRUSR | S_IWUSR));
+	umask(om);
+
+	if (ret < 0)
+		return -1;
+        return open(path, O_RDONLY | O_NONBLOCK);
+}
+
+static int check_override(int ofd, char *nodename, int timeout)
+{
+	char buf[128];
+	fd_set rfds;
+	struct timeval tv = {0, 0};
+	int ret, x;
+
+	if (ofd < 0 || !nodename || !strlen(nodename)) {
+		sleep(timeout);
+		return 0;
+	}
+
+	FD_ZERO(&rfds);
+	FD_SET(ofd, &rfds);
+	tv.tv_usec = 0;
+	tv.tv_sec = timeout;
+
+	ret = select(ofd + 1, &rfds, NULL, NULL, &tv);
+	if (ret < 0) {
+		syslog(LOG_ERR, "select: %s\n", strerror(errno));
+		return -1;
+	}
+
+	if (ret == 0)
+		return 0;
+
+	memset(buf, 0, sizeof(buf));
+	ret = read(ofd, buf, sizeof(buf) - 1);
+	if (ret < 0) {
+		syslog(LOG_ERR, "read: %s\n", strerror(errno));
+		return -1;
+	}
+
+	/* chop off control characters */
+	for (x = 0; x < ret; x++) {
+		if (buf[x] < 0x20) {
+			buf[x] = 0;
+			break;
+		}
+	}
+
+	if (!strcasecmp(nodename, buf)) {
+		/* Case insensitive, but not as nice as, say, name_equal
+		   in the other file... */
+		return 1;
+	}
+
+	return 0;
+}
+
+
 /* If there are victims after a node has joined, it's a good indication that
    they may be joining the cluster shortly.  If we delay a bit they might
    become members and we can avoid fencing them.  This is only really an issue
@@ -282,6 +358,7 @@
 	fd_node_t *node;
 	char *master_name;
 	int master, error, cd;
+	int override = -1;
 
 	master = find_master_nodeid(fd, &master_name);
 
@@ -318,7 +395,22 @@
 			list_del(&node->list);
 			free(node);
 		}
-		sleep(5);
+
+		if (!comline.override_path) {
+			sleep(5);
+			continue;
+		}
+
+		/* Check for manual intervention */
+		override = open_override(comline.override_path);
+		if (check_override(override, node->name, 5) > 0) {
+			syslog(LOG_WARNING, "fence \"%s\" overridden by "
+			       "administrator intervention", node->name);
+
+			list_del(&node->list);
+			free(node);
+		}
+		close_override(&override, comline.override_path);
 	}
 
 	ccs_disconnect(cd);



             reply	other threads:[~2007-01-29 20:30 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-01-29 20:30 lhh [this message]
  -- strict thread matches above, loose matches on Subject: below --
2007-01-29 20:04 [Cluster-devel] cluster/fence/fenced fd.h main.c recover.c lhh
2007-01-29 19:55 lhh

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20070129203026.13144.qmail@sourceware.org \
    --to=lhh@sourceware.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.