dlm: master - dlm_tool: update man page
by David Teigland
Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=...
Commit: adff37be43519da0cf40d7de541d217fbcf842c5
Parent: 878c9fa5cd0d9a8be0f6e65361f50f43a8ec4843
Author: David Teigland <teigland(a)redhat.com>
AuthorDate: Thu Apr 5 15:13:07 2012 -0500
Committer: David Teigland <teigland(a)redhat.com>
CommitterDate: Thu Apr 5 15:13:07 2012 -0500
dlm_tool: update man page
and help output
Signed-off-by: David Teigland <teigland(a)redhat.com>
---
dlm_tool/dlm_tool.8 | 105 +++++++++++++++++++++++++--------------------------
dlm_tool/main.c | 22 +++++-----
2 files changed, 63 insertions(+), 64 deletions(-)
diff --git a/dlm_tool/dlm_tool.8 b/dlm_tool/dlm_tool.8
index df9aa64..9b061a0 100644
--- a/dlm_tool/dlm_tool.8
+++ b/dlm_tool/dlm_tool.8
@@ -1,4 +1,4 @@
-.TH DLM_TOOL 8 2009-01-20 cluster cluster
+.TH DLM_TOOL 8 2012-04-05 dlm dlm
.SH NAME
dlm_tool \- a utility for the dlm and dlm_controld daemon
@@ -10,89 +10,88 @@ dlm_tool \- a utility for the dlm and dlm_controld daemon
.I name
]
-.SH DESCRIPTION
-.TP
+.SH COMMANDS
+
.B ls
-Display internal dlm_controld state about lockspaces.
+.br
+ Display dlm_controld internal lockspace state.
+
+.B status
+.br
+ Dump dlm_controld daemon state.
-.TP
.B dump
-Dump dlm_controld debug buffer.
+.br
+ Dump dlm_controld debug buffer.
-.TP
-.B log_plock
-Dump dlm_controld plock debug buffer.
+.B dump_config
+.br
+ Dump dlm_controld config settings.
-.TP
-.BI plocks " name"
-Dump posix locks from dlm_controld for the lockspace.
+.BI fence_ack " nodeid"
+.br
+ Quit trying to fence a node.
-.TP
-.BI lockdump " name"
-Minimal display of locks from the lockspace.
+.B log_plock
+.br
+ Dump dlm_controld plock debug buffer.
-.TP
-.BI lockdebug " name"
-Extended display of locks from the lockspace.
+.BI plocks " name"
+.br
+ Dump posix locks from dlm_controld for the lockspace.
-.TP
.BI join " name"
-Join a lockspace.
+.br
+ Join a lockspace.
-.TP
.BI leave " name"
-Leave a lockspace.
+.br
+ Leave a lockspace.
-.TP
-.BI deadlock_check " name"
-Start a deadlock detection cycle for the lockspace.
+.BI lockdebug " name"
+.br
+ Complete display of locks from the lockspace.
+
+.BI lockdump " name"
+.br
+ Minimal display of locks from the lockspace (deprecated).
.SH OPTIONS
-.TP
+
.B \-n
Show all node information in ls.
-.TP
-.BI \-d " num"
-Resource directory enabled (1) or disabled (0) during join. Default 0.
+.B \-d
+0|1 Resource directory off/on in join, default 0
-.TP
-.BI \-e " num"
-Exclusive create off/on (0/1) in join. Default 0.
+.B \-e
+0|1 Exclusive create off/on in join, default 0
-.TP
-.BI \-f " num"
-FS memory allocation off/on (0/1) in join. Default 0.
+.B \-f
+0|1 FS (filesystem) flag off/on in join, default 0
-.TP
.BI \-m " mode"
-The permission mode (in octal) of the lockspace device created by join.
-Default 0600.
+Permission mode for lockspace device (octal), default 0600
-.TP
-.B \-M
-Dump MSTCPY locks in addition to locks held by local processes.
-
-.TP
.B \-s
-Summary following lockdebug output (experimental, format may change).
+Summary following lockdebug output (experiemental)
-.TP
.B \-v
-Verbose lockdebug output.
+Verbose lockdebug output
-.TP
.B \-w
-Wide lockdebug output.
+Wide lockdebug output
+
+.B \-M
+Include MSTCPY locks in lockdump output
-.TP
.B \-h
-Print a help message describing available options, then exit.
+Print help, then exit
-.TP
.B \-V
-Print program version information, then exit.
+Print program version information, then exit
.SH SEE ALSO
-.BR dlm_controld (8)
+.BR dlm_controld (8),
+.BR dlm.conf (5)
diff --git a/dlm_tool/main.c b/dlm_tool/main.c
index 1fb6b3e..8dba291 100644
--- a/dlm_tool/main.c
+++ b/dlm_tool/main.c
@@ -183,23 +183,23 @@ static void print_usage(void)
{
printf("Usage:\n");
printf("\n");
- printf("dlm_tool [options] [join | leave | lockdump | lockdebug |\n"
- " ls | dump | dump_config | log_plock | plocks |\n"
- " fence_ack]\n");
+ printf("dlm_tool [command] [options] [name]\n");
+ printf("\n");
+ printf("Commands:\n");
+ printf("ls, status, dump, dump_config, fence_ack\n");
+ printf("log_plock, plocks\n");
+ printf("join, leave, lockdebug\n");
printf("\n");
printf("Options:\n");
printf(" -n Show all node information in ls\n");
- printf(" -d <n> Resource directory off/on (0/1) in join, default 0\n");
- printf(" -e <n> Exclusive create off/on (0/1) in join, default 0\n");
- printf(" -f <n> FS memory allocation off/on (0/1) in join, default 0\n");
+ printf(" -d 0|1 Resource directory off/on in join, default 0\n");
+ printf(" -e 0|1 Exclusive create off/on in join, default 0\n");
+ printf(" -f 0|1 FS (filesystem) flag off/on in join, default 0\n");
printf(" -m <mode> Permission mode for lockspace device (octal), default 0600\n");
- printf(" -M Print MSTCPY locks in lockdump\n"
- " (remote locks that are locally mastered)\n");
- printf(" -s Summary following lockdebug output\n");
- printf(" (experimental, format not fixed)\n");
+ printf(" -s Summary following lockdebug output (experimental)\n");
printf(" -v Verbose lockdebug output\n");
printf(" -w Wide lockdebug output\n");
- printf(" -h Print this help, then exit\n");
+ printf(" -h Print help, then exit\n");
printf(" -V Print program version information, then exit\n");
printf("\n");
}
12 years, 1 month
- 1
- 0
dlm: master - dlm_controld: new config handling
by David Teigland
Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=...
Commit: ff1f5c78203613a8e28be6d7de1d558d598f8ca9
Parent: b8d1a03c2d23a42761216ac33dae6e42f87ec782
Author: David Teigland <teigland(a)redhat.com>
AuthorDate: Thu Apr 5 13:47:23 2012 -0500
Committer: David Teigland <teigland(a)redhat.com>
CommitterDate: Thu Apr 5 13:47:23 2012 -0500
dlm_controld: new config handling
to handle both command line and config file options.
Signed-off-by: David Teigland <teigland(a)redhat.com>
---
dlm_controld/action.c | 45 ++--
dlm_controld/config.c | 127 ++++++-----
dlm_controld/cpg.c | 54 +++---
dlm_controld/daemon_cpg.c | 28 ++-
dlm_controld/dlm_controld.h | 1 +
dlm_controld/dlm_daemon.h | 119 +++++-----
dlm_controld/lib.c | 5 +
dlm_controld/libdlmcontrol.h | 1 +
dlm_controld/logging.c | 6 +-
dlm_controld/main.c | 525 +++++++++++++++++++++++++++++------------
dlm_controld/member.c | 2 +-
dlm_controld/plock.c | 42 ++--
dlm_tool/main.c | 22 ++-
13 files changed, 615 insertions(+), 362 deletions(-)
diff --git a/dlm_controld/action.c b/dlm_controld/action.c
index 6cb34f2..8ac05b6 100644
--- a/dlm_controld/action.c
+++ b/dlm_controld/action.c
@@ -50,9 +50,6 @@ static int detect_protocol(void)
return proto;
}
-/* TODO: getting cluster_name will almost certainly change to either
- a corosync lib api, or some cmap path other than cluster.name */
-
static int detect_cluster_name(void)
{
cmap_handle_t handle;
@@ -799,7 +796,8 @@ void clear_configfs(void)
int setup_configfs_options(void)
{
- int rv;
+ char *proto_name;
+ int rv, proto_num;
clear_configfs();
@@ -808,24 +806,33 @@ int setup_configfs_options(void)
return rv;
/* the kernel has its own defaults for these values which we
- don't want to change unless these have been set; -1 means
- they have not been set on command line or config file */
+ don't want to change unless these have been set explicitly
+ on cli or config file */
- if (cfgk_debug != -1)
- set_configfs_cluster("log_debug", NULL, cfgk_debug);
- if (cfgk_timewarn != -1)
- set_configfs_cluster("timewarn_cs", NULL, cfgk_timewarn);
+ if (dlm_options[log_debug_ind].cli_set ||
+ dlm_options[log_debug_ind].file_set)
+ set_configfs_cluster("log_debug", NULL, opt(log_debug_ind));
- if (cfgk_protocol == PROTO_DETECT) {
- rv = detect_protocol();
- if (rv == PROTO_TCP || rv == PROTO_SCTP)
- cfgk_protocol = rv;
- }
+ if (dlm_options[timewarn_ind].cli_set ||
+ dlm_options[timewarn_ind].file_set)
+ set_configfs_cluster("timewarn_cs", NULL, opt(timewarn_ind));
+
+ proto_name = opts(protocol_ind);
+ proto_num = -1;
+
+ if (!strcasecmp(proto_name, "detect") || !strcmp(proto_name, "2"))
+ proto_num = detect_protocol(); /* may be -1 */
+
+ else if (!strcasecmp(proto_name, "tcp") || !strcmp(proto_name, "0"))
+ proto_num = PROTO_TCP;
+
+ else if (!strcasecmp(proto_name, "sctp") || !strcmp(proto_name, "1"))
+ proto_num = PROTO_SCTP;
- if (cfgk_protocol == PROTO_TCP || cfgk_protocol == PROTO_SCTP)
- set_configfs_cluster("protocol", NULL, cfgk_protocol);
+ if (proto_num == PROTO_TCP || proto_num == PROTO_SCTP)
+ set_configfs_cluster("protocol", NULL, proto_num);
- if (cfgk_protocol == PROTO_SCTP)
+ if (proto_num == PROTO_SCTP)
set_proc_rmem();
/*
@@ -838,7 +845,7 @@ int setup_configfs_options(void)
* go ahead without the monitor being open
*/
- if (cfgd_enable_fscontrol) {
+ if (opt(enable_fscontrol_ind)) {
/* deprecated */
set_configfs_cluster("recover_callbacks", NULL, 0);
} else {
diff --git a/dlm_controld/config.c b/dlm_controld/config.c
index 040784b..1efea45 100644
--- a/dlm_controld/config.c
+++ b/dlm_controld/config.c
@@ -8,14 +8,18 @@
#include "dlm_daemon.h"
-/* TODO:
- <dlm>
- <lockspace name="foo" nodir="1">
- <master nodeid="1" weight="2"/>
- <master nodeid="2" weight="3"/>
- </lockspace>
- </dlm>
-*/
+/*
+ * TODO: lockspace master/nodir/weight
+ *
+ * lockspace foo nodir=1
+ * master foo nodeid=1 weight=1
+ * master foo nodeid=2 weight=1
+ * master foo nodeid=3 weight=1
+ *
+ * lockspace bar nodir=1
+ * master bar nodeid=4 weight=2
+ * master bar nodeid=5 weight=1
+ */
int get_weight(int nodeid, char *lockspace)
{
@@ -23,20 +27,7 @@ int get_weight(int nodeid, char *lockspace)
return 1;
}
-static void proto_val(char *str, int *val)
-{
- if (!strncasecmp(str, "tcp", 3))
- *val = PROTO_TCP;
- else if (!strncasecmp(str, "sctp", 4))
- *val = PROTO_SCTP;
- else if (!strncasecmp(str, "detect", 6))
- *val = PROTO_DETECT;
- else {
- log_error("invalid protocol value %s", str);
- }
-}
-
-static void set_val(char *line, int *val_out)
+static void get_val_int(char *line, int *val_out)
{
char key[PATH_MAX];
char val[PATH_MAX];
@@ -47,11 +38,9 @@ static void set_val(char *line, int *val_out)
return;
*val_out = atoi(val);
-
- log_debug("config %s=%d", key, *val_out);
}
-static void get_val(char *line, char *val_out)
+static void get_val_str(char *line, char *val_out)
{
char key[PATH_MAX];
char val[PATH_MAX];
@@ -66,9 +55,11 @@ static void get_val(char *line, char *val_out)
void setup_config(int update)
{
+ struct dlm_option *o;
FILE *file;
char line[PATH_MAX];
char str[PATH_MAX];
+ int i, val;
if (!path_exists(CONF_FILE_PATH))
return;
@@ -83,59 +74,71 @@ void setup_config(int update)
if (line[0] == '\n')
continue;
- if (!optk_debug && !strncmp(line, "log_debug", strlen("log_debug")))
- set_val(line, &cfgk_debug);
+ memset(str, 0, sizeof(str));
+
+ for (i = 0; i < PATH_MAX; i++) {
+ if (line[i] == ' ')
+ break;
+ if (line[i] == '=')
+ break;
+ if (line[i] == '\0')
+ break;
+ if (line[i] == '\n')
+ break;
+ if (line[i] == '\t')
+ break;
+ str[i] = line[i];
+ }
- else if (!optk_timewarn && !strncmp(line, "timewarn", strlen("timewarn")) && !update)
- set_val(line, &cfgk_timewarn);
+ o = get_dlm_option(str);
+ if (!o)
+ continue;
- else if (!optd_post_join_delay && !strncmp(line, "post_join_delay", strlen("post_join_delay")))
- set_val(line, &cfgd_post_join_delay);
+ o->file_set++;
- else if (!optd_enable_fencing && !strncmp(line, "enable_fencing", strlen("enable_fencing")) && !update)
- set_val(line, &cfgd_enable_fencing);
+ if (!o->req_arg) {
+ /* ignore any = x */
- else if (!optd_enable_startup_fencing && !strncmp(line, "enable_startup_fencing", strlen("enable_startup_fencing")) && !update)
- set_val(line, &cfgd_enable_startup_fencing);
+ o->file_int = 1;
- else if (!optd_enable_concurrent_fencing && !strncmp(line, "enable_concurrent_fencing", strlen("enable_concurrent_fencing")) && !update)
- set_val(line, &cfgd_enable_concurrent_fencing);
+ if (!o->cli_set)
+ o->use_int = o->file_int;
- else if (!optd_enable_quorum_fencing && !strncmp(line, "enable_quorum_fencing", strlen("enable_quorum_fencing")) && !update)
- set_val(line, &cfgd_enable_quorum_fencing);
+ log_debug("config file %s = %d cli_set %d use %d",
+ o->name, o->file_int, o->cli_set, o->use_int);
- else if (!optd_enable_quorum_lockspace && !strncmp(line, "enable_quorum_lockspace", strlen("enable_quorum_lockspace")) && !update)
- set_val(line, &cfgd_enable_quorum_lockspace);
+ } else if (o->req_arg == req_arg_int) {
+ get_val_int(line, &val);
- else if (!optd_enable_fscontrol && !strncmp(line, "enable_fscontrol", strlen("enable_fscontrol")) && !update)
- set_val(line, &cfgd_enable_fscontrol);
+ o->file_int = val;
- else if (!optd_enable_plock && !strncmp(line, "enable_plock", strlen("enable_plock")) && !update)
- set_val(line, &cfgd_enable_plock);
+ if (!o->cli_set)
+ o->use_int = o->file_int;
- else if (!optd_plock_ownership && !strncmp(line, "plock_ownership", strlen("plock_ownership")) && !update)
- set_val(line, &cfgd_plock_ownership);
+ log_debug("config file %s = %d cli_set %d use %d",
+ o->name, o->file_int, o->cli_set, o->use_int);
- else if (!optd_plock_debug && !strncmp(line, "plock_debug", strlen("plock_debug")))
- set_val(line, &cfgd_plock_debug);
+ } else if (o->req_arg == req_arg_bool) {
+ get_val_int(line, &val);
- else if (!optd_plock_rate_limit && !strncmp(line, "plock_rate_limit", strlen("plock_rate_limit")))
- set_val(line, &cfgd_plock_rate_limit);
+ o->file_int = val ? 1 : 0;
- else if (!optd_drop_resources_time && !strncmp(line, "drop_resources_time", strlen("drop_resources_time")))
- set_val(line, &cfgd_drop_resources_time);
+ if (!o->cli_set)
+ o->use_int = o->file_int;
- else if (!optd_drop_resources_count && !strncmp(line, "drop_resources_count", strlen("drop_resources_count")))
- set_val(line, &cfgd_drop_resources_count);
+ log_debug("config file %s = %d cli_set %d use %d",
+ o->name, o->file_int, o->cli_set, o->use_int);
+ } else if (o->req_arg == req_arg_str) {
+ memset(str, 0, sizeof(str));
+ get_val_str(line, str);
- else if (!optd_drop_resources_age && !strncmp(line, "drop_resources_age", strlen("drop_resources_age")))
- set_val(line, &cfgd_drop_resources_age);
+ o->file_str = strdup(str);
- else if (!optk_protocol && !strncmp(line, "protocol", strlen("protocol")) && !update) {
- memset(str, 0, sizeof(str));
- get_val(line, str);
- proto_val(str, &cfgk_protocol);
- log_debug("config protocol = %d", cfgk_protocol);
+ if (!o->cli_set)
+ o->use_str = o->file_str;
+
+ log_debug("config file %s = %s cli_set %d use %s",
+ o->name, o->file_str, o->cli_set, o->use_str);
}
}
diff --git a/dlm_controld/cpg.c b/dlm_controld/cpg.c
index 83e9b17..49fb06a 100644
--- a/dlm_controld/cpg.c
+++ b/dlm_controld/cpg.c
@@ -315,7 +315,7 @@ static void node_history_lockspace_fail(struct lockspace *ls, int nodeid,
return;
}
- if (cfgd_enable_fencing && node->start_time) {
+ if (opt(enable_fencing_ind) && node->start_time) {
node->need_fencing = 1;
node->fence_queries = 0;
}
@@ -396,7 +396,7 @@ static int check_fencing_done(struct lockspace *ls)
int wait_count = 0;
int rv, in_progress;
- if (!cfgd_enable_fencing) {
+ if (!opt(enable_fencing_ind)) {
log_group(ls, "check_fencing disabled");
return 1;
}
@@ -592,7 +592,7 @@ static int wait_conditions_done(struct lockspace *ls)
return 0;
}
- if ((cfgd_enable_quorum_lockspace > 1) && !cluster_quorate) {
+ if (opt(enable_quorum_lockspace_ind) && !cluster_quorate) {
log_group(ls, "wait for quorum");
ls->wait_debug = DLMC_LS_WAIT_QUORUM;
return 0;
@@ -1097,7 +1097,7 @@ static void prepare_plocks(struct lockspace *ls)
struct member *memb;
uint32_t plocks_data;
- if (!cfgd_enable_plock || ls->disable_plock)
+ if (!opt(enable_plock_ind) || ls->disable_plock)
return;
log_dlock(ls, "prepare_plocks");
@@ -1393,6 +1393,9 @@ static void deliver_cb(cpg_handle_t handle,
int ignore_plock;
int rv;
+ int enable_plock = opt(enable_plock_ind);
+ int plock_ownership = opt(plock_ownership_ind);
+
ls = find_ls_handle(handle);
if (!ls) {
log_error("deliver_cb no ls for cpg %s", group_name->value);
@@ -1425,11 +1428,11 @@ static void deliver_cb(cpg_handle_t handle,
ignore_plock = 1;
break;
}
- if (cfgd_enable_plock)
+ if (enable_plock)
receive_plock(ls, hd, len);
else
log_error("msg %d nodeid %d enable_plock %d",
- hd->type, nodeid, cfgd_enable_plock);
+ hd->type, nodeid, enable_plock);
break;
case DLM_MSG_PLOCK_OWN:
@@ -1439,12 +1442,11 @@ static void deliver_cb(cpg_handle_t handle,
ignore_plock = 1;
break;
}
- if (cfgd_enable_plock && cfgd_plock_ownership)
+ if (enable_plock && plock_ownership)
receive_own(ls, hd, len);
else
log_error("msg %d nodeid %d enable_plock %d owner %d",
- hd->type, nodeid, cfgd_enable_plock,
- cfgd_plock_ownership);
+ hd->type, nodeid, enable_plock, plock_ownership);
break;
case DLM_MSG_PLOCK_DROP:
@@ -1454,12 +1456,11 @@ static void deliver_cb(cpg_handle_t handle,
ignore_plock = 1;
break;
}
- if (cfgd_enable_plock && cfgd_plock_ownership)
+ if (enable_plock && plock_ownership)
receive_drop(ls, hd, len);
else
log_error("msg %d nodeid %d enable_plock %d owner %d",
- hd->type, nodeid, cfgd_enable_plock,
- cfgd_plock_ownership);
+ hd->type, nodeid, enable_plock, plock_ownership);
break;
case DLM_MSG_PLOCK_SYNC_LOCK:
@@ -1470,65 +1471,64 @@ static void deliver_cb(cpg_handle_t handle,
ignore_plock = 1;
break;
}
- if (cfgd_enable_plock && cfgd_plock_ownership)
+ if (enable_plock && plock_ownership)
receive_sync(ls, hd, len);
else
log_error("msg %d nodeid %d enable_plock %d owner %d",
- hd->type, nodeid, cfgd_enable_plock,
- cfgd_plock_ownership);
+ hd->type, nodeid, enable_plock, plock_ownership);
break;
case DLM_MSG_PLOCKS_DATA:
if (ls->disable_plock)
break;
- if (cfgd_enable_plock)
+ if (enable_plock)
receive_plocks_data(ls, hd, len);
else
log_error("msg %d nodeid %d enable_plock %d",
- hd->type, nodeid, cfgd_enable_plock);
+ hd->type, nodeid, enable_plock);
break;
case DLM_MSG_PLOCKS_DONE:
if (ls->disable_plock)
break;
- if (cfgd_enable_plock)
+ if (enable_plock)
receive_plocks_done(ls, hd, len);
else
log_error("msg %d nodeid %d enable_plock %d",
- hd->type, nodeid, cfgd_enable_plock);
+ hd->type, nodeid, enable_plock);
break;
#if 0
case DLM_MSG_DEADLK_CYCLE_START:
- if (cfgd_enable_deadlk)
+ if (opt(enable_deadlk))
receive_cycle_start(ls, hd, len);
else
log_error("msg %d nodeid %d enable_deadlk %d",
- hd->type, nodeid, cfgd_enable_deadlk);
+ hd->type, nodeid, opt(enable_deadlk));
break;
case DLM_MSG_DEADLK_CYCLE_END:
- if (cfgd_enable_deadlk)
+ if (opt(enable_deadlk))
receive_cycle_end(ls, hd, len);
else
log_error("msg %d nodeid %d enable_deadlk %d",
- hd->type, nodeid, cfgd_enable_deadlk);
+ hd->type, nodeid, opt(enable_deadlk));
break;
case DLM_MSG_DEADLK_CHECKPOINT_READY:
- if (cfgd_enable_deadlk)
+ if (opt(enable_deadlk))
receive_checkpoint_ready(ls, hd, len);
else
log_error("msg %d nodeid %d enable_deadlk %d",
- hd->type, nodeid, cfgd_enable_deadlk);
+ hd->type, nodeid, opt(enable_deadlk));
break;
case DLM_MSG_DEADLK_CANCEL_LOCK:
- if (cfgd_enable_deadlk)
+ if (opt(enable_deadlk))
receive_cancel_lock(ls, hd, len);
else
log_error("msg %d nodeid %d enable_deadlk %d",
- hd->type, nodeid, cfgd_enable_deadlk);
+ hd->type, nodeid, opt(enable_deadlk));
break;
#endif
diff --git a/dlm_controld/daemon_cpg.c b/dlm_controld/daemon_cpg.c
index a68de80..3e7ed74 100644
--- a/dlm_controld/daemon_cpg.c
+++ b/dlm_controld/daemon_cpg.c
@@ -457,7 +457,11 @@ static struct node_daemon *add_node_daemon(int nodeid)
/* explicit command line arg has first priority */
- if (optd_fence_all_agent) {
+ if (dlm_options[fence_all_ind].cli_set) {
+ memset(&fence_all_device, 0, sizeof(struct fence_device));
+ strcpy(fence_all_device.name, "fence_all");
+ strcpy(fence_all_device.agent, dlm_options[fence_all_ind].cli_str);
+
fc->dev[0] = &fence_all_device;
goto out;
}
@@ -746,7 +750,7 @@ static void daemon_fence_work(void)
/* poll_fencing++; */
- if (cfgd_enable_quorum_fencing && !cluster_quorate) {
+ if (opt(enable_quorum_fencing_ind) && !cluster_quorate) {
/* wait for quorum before doing any fencing, but if there
is none, send_fence_clear below can unblock new nodes */
log_debug("fence work wait for quorum");
@@ -766,12 +770,12 @@ static void daemon_fence_work(void)
continue;
}
- if (!cfgd_enable_startup_fencing)
+ if (!opt(enable_startup_fencing_ind))
continue;
- if (monotime() - daemon_last_join_monotime < cfgd_post_join_delay) {
+ if (monotime() - daemon_last_join_monotime < opt(post_join_delay_ind)) {
log_debug("fence startup %d delay %d from %llu",
- node->nodeid, cfgd_post_join_delay,
+ node->nodeid, opt(post_join_delay_ind),
(unsigned long long)daemon_last_join_monotime);
poll_fencing++;
continue;
@@ -842,7 +846,7 @@ static void daemon_fence_work(void)
continue;
}
- if (!cfgd_enable_concurrent_fencing && daemon_fence_pid) {
+ if (!opt(enable_concurrent_fencing_ind) && daemon_fence_pid) {
/* run one agent at a time in case they need the same switch */
log_debug("fence request %d delay for other pid %d",
node->nodeid, daemon_fence_pid);
@@ -851,9 +855,9 @@ static void daemon_fence_work(void)
continue;
}
- if (monotime() - cluster_last_join_monotime < cfgd_post_join_delay) {
+ if (monotime() - cluster_last_join_monotime < opt(post_join_delay_ind)) {
log_debug("fence request %d delay %d from %llu",
- node->nodeid, cfgd_post_join_delay,
+ node->nodeid, opt(post_join_delay_ind),
(unsigned long long)cluster_last_join_monotime);
node->delay_fencing = 1;
poll_fencing++;
@@ -1000,7 +1004,7 @@ static void daemon_fence_work(void)
* clear fence_in_progress_unknown
*/
out_fipu:
- if (cfgd_enable_startup_fencing &&
+ if (opt(enable_startup_fencing_ind) &&
fence_in_progress_unknown &&
list_empty(&startup_nodes) &&
!wait_clear_fipu &&
@@ -1063,7 +1067,7 @@ static void daemon_fence_work(void)
}
}
- if (!cfgd_enable_startup_fencing && fence_in_progress_unknown) {
+ if (!opt(enable_startup_fencing_ind) && fence_in_progress_unknown) {
/*
* case C in comment above
* all nodes are starting and have fipu set. All expect a
@@ -1872,7 +1876,7 @@ static void confchg_cb_daemon(cpg_handle_t handle,
reason = 0;
low = 0;
- if (!cfgd_enable_fencing)
+ if (!opt(enable_fencing_ind))
continue;
if (node->need_fencing) {
@@ -1959,7 +1963,7 @@ int setup_cpg_daemon(void)
memset(&our_protocol, 0, sizeof(our_protocol));
- if (cfgd_enable_fscontrol)
+ if (opt(enable_fscontrol_ind))
our_protocol.daemon_max[0] = 2;
else
our_protocol.daemon_max[0] = 3;
diff --git a/dlm_controld/dlm_controld.h b/dlm_controld/dlm_controld.h
index 84662eb..334408f 100644
--- a/dlm_controld/dlm_controld.h
+++ b/dlm_controld/dlm_controld.h
@@ -31,6 +31,7 @@
#define DLMC_CMD_DUMP_LOG_PLOCK 11
#define DLMC_CMD_FENCE_ACK 12
#define DLMC_CMD_DUMP_STATUS 13
+#define DLMC_CMD_DUMP_CONFIG 14
struct dlmc_header {
unsigned int magic;
diff --git a/dlm_controld/dlm_daemon.h b/dlm_controld/dlm_daemon.h
index 4f3e551..788f4e8 100644
--- a/dlm_controld/dlm_daemon.h
+++ b/dlm_controld/dlm_daemon.h
@@ -80,22 +80,65 @@
#define DEFAULT_LOGFILE_PRIORITY LOG_INFO
#define DEFAULT_LOGFILE LOG_FILE_PATH
-#define DEFAULT_DEBUG_LOGFILE 0
-#define DEFAULT_POST_JOIN_DELAY 30
-#define DEFAULT_ENABLE_FENCING 1
-#define DEFAULT_ENABLE_STARTUP_FENCING 1
-#define DEFAULT_ENABLE_CONCURRENT_FENCING 0
-#define DEFAULT_ENABLE_QUORUM_FENCING 1
-#define DEFAULT_ENABLE_QUORUM_LOCKSPACE 0
-#define DEFAULT_ENABLE_FSCONTROL 0
-#define DEFAULT_ENABLE_PLOCK 1
-#define DEFAULT_PLOCK_DEBUG 0
-#define DEFAULT_PLOCK_RATE_LIMIT 0
-#define DEFAULT_PLOCK_OWNERSHIP 0
-#define DEFAULT_DROP_RESOURCES_TIME 10000 /* 10 sec */
-#define DEFAULT_DROP_RESOURCES_COUNT 10
-#define DEFAULT_DROP_RESOURCES_AGE 10000 /* 10 sec */
-#define DEFAULT_FENCE_ALL_AGENT "dlm_stonith"
+enum {
+ no_arg = 0,
+ req_arg_bool = 1,
+ req_arg_int = 2,
+ req_arg_str = 3,
+};
+
+enum {
+ daemon_debug_ind = 0,
+ log_debug_ind,
+ timewarn_ind,
+ protocol_ind,
+ debug_logfile_ind,
+ enable_fscontrol_ind,
+ enable_plock_ind,
+ plock_debug_ind,
+ plock_rate_limit_ind,
+ plock_ownership_ind,
+ drop_resources_time_ind,
+ drop_resources_count_ind,
+ drop_resources_age_ind,
+ post_join_delay_ind,
+ enable_fencing_ind,
+ enable_concurrent_fencing_ind,
+ enable_startup_fencing_ind,
+ enable_quorum_fencing_ind,
+ enable_quorum_lockspace_ind,
+ fence_all_ind,
+ unfence_all_ind,
+ help_ind,
+ version_ind,
+ dlm_options_max,
+};
+
+struct dlm_option {
+ const char *name;
+ char letter;
+ int req_arg;
+ const char *desc;
+
+ int use_int;
+ char *use_str;
+
+ int default_int;
+ const char *default_str;
+
+ int cli_set;
+ int cli_int;
+ char *cli_str;
+
+ int file_set;
+ int file_int;
+ char *file_str;
+};
+
+EXTERN struct dlm_option dlm_options[dlm_options_max];
+#define opt(x) dlm_options[x].use_int
+#define opts(x) dlm_options[x].use_str
+
/* DLM_LOCKSPACE_LEN: maximum lockspace name length, from linux/dlmconstants.h.
Copied in libdlm.h so apps don't need to include the kernel header.
@@ -116,13 +159,10 @@
#define MAXLINE 256
-/* cfgk_protocol */
-
#define PROTO_TCP 0
#define PROTO_SCTP 1
#define PROTO_DETECT 2
-EXTERN int daemon_debug_opt;
EXTERN int daemon_quit;
EXTERN int cluster_down;
EXTERN int poll_lockspaces;
@@ -146,46 +186,6 @@ EXTERN uint32_t monitor_minor;
EXTERN uint32_t plock_minor;
EXTERN struct fence_device fence_all_device;
-EXTERN int optk_debug;
-EXTERN int optk_timewarn;
-EXTERN int optk_protocol;
-EXTERN int optd_debug_logfile;
-EXTERN int optd_post_join_delay;
-EXTERN int optd_enable_fencing;
-EXTERN int optd_enable_startup_fencing;
-EXTERN int optd_enable_concurrent_fencing;
-EXTERN int optd_enable_quorum_fencing;
-EXTERN int optd_enable_quorum_lockspace;
-EXTERN int optd_enable_fscontrol;
-EXTERN int optd_enable_plock;
-EXTERN int optd_plock_debug;
-EXTERN int optd_plock_rate_limit;
-EXTERN int optd_plock_ownership;
-EXTERN int optd_drop_resources_time;
-EXTERN int optd_drop_resources_count;
-EXTERN int optd_drop_resources_age;
-EXTERN int optd_fence_all_agent;
-
-EXTERN int cfgk_debug;
-EXTERN int cfgk_timewarn;
-EXTERN int cfgk_protocol;
-EXTERN int cfgd_debug_logfile;
-EXTERN int cfgd_post_join_delay;
-EXTERN int cfgd_enable_fencing;
-EXTERN int cfgd_enable_startup_fencing;
-EXTERN int cfgd_enable_concurrent_fencing;
-EXTERN int cfgd_enable_quorum_fencing;
-EXTERN int cfgd_enable_quorum_lockspace;
-EXTERN int cfgd_enable_fscontrol;
-EXTERN int cfgd_enable_plock;
-EXTERN int cfgd_plock_debug;
-EXTERN int cfgd_plock_rate_limit;
-EXTERN int cfgd_plock_ownership;
-EXTERN int cfgd_drop_resources_time;
-EXTERN int cfgd_drop_resources_count;
-EXTERN int cfgd_drop_resources_age;
-EXTERN char fence_all_agent[PATH_MAX];
-
#define LOG_DUMP_SIZE DLMC_DUMP_SIZE
#define LOG_PLOCK 0x00010000
@@ -384,6 +384,7 @@ struct lockspace *find_ls(char *name);
struct lockspace *find_ls_id(uint32_t id);
const char *dlm_mode_str(int mode);
void cluster_dead(int ci);
+struct dlm_option *get_dlm_option(char *name);
/* member.c */
int setup_cluster(void);
diff --git a/dlm_controld/lib.c b/dlm_controld/lib.c
index a7cd443..8b079b1 100644
--- a/dlm_controld/lib.c
+++ b/dlm_controld/lib.c
@@ -142,6 +142,11 @@ int dlmc_dump_debug(char *buf)
return do_dump(DLMC_CMD_DUMP_DEBUG, NULL, buf);
}
+int dlmc_dump_config(char *buf)
+{
+ return do_dump(DLMC_CMD_DUMP_CONFIG, NULL, buf);
+}
+
int dlmc_dump_log_plock(char *buf)
{
return do_dump(DLMC_CMD_DUMP_LOG_PLOCK, NULL, buf);
diff --git a/dlm_controld/libdlmcontrol.h b/dlm_controld/libdlmcontrol.h
index 2af30de..436088b 100644
--- a/dlm_controld/libdlmcontrol.h
+++ b/dlm_controld/libdlmcontrol.h
@@ -79,6 +79,7 @@ struct dlmc_lockspace {
#define DLMC_NODES_NEXT 3
int dlmc_dump_debug(char *buf);
+int dlmc_dump_config(char *buf);
int dlmc_dump_log_plock(char *buf);
int dlmc_dump_plocks(char *name, char *buf);
int dlmc_lockspace_info(char *lsname, struct dlmc_lockspace *ls);
diff --git a/dlm_controld/logging.c b/dlm_controld/logging.c
index 1f97df7..1b08c4f 100644
--- a/dlm_controld/logging.c
+++ b/dlm_controld/logging.c
@@ -24,7 +24,7 @@ void init_logging(void)
/* logfile_priority is the only one of these options that
can be controlled from command line or environment variable */
- if (cfgd_debug_logfile)
+ if (opt(debug_logfile_ind))
logfile_priority = LOG_DEBUG;
if (logfile[0]) {
@@ -167,10 +167,10 @@ void log_level(char *name_in, uint32_t level_in, const char *fmt, ...)
fflush(logfile_fp);
}
- if (!daemon_debug_opt)
+ if (!dlm_options[daemon_debug_ind].use_int)
return;
- if (level || (plock && cfgd_plock_debug))
+ if (level || (plock && opt(plock_debug_ind)))
fprintf(stderr, "%s", log_str);
}
diff --git a/dlm_controld/main.c b/dlm_controld/main.c
index dcb01ec..872cdfd 100644
--- a/dlm_controld/main.c
+++ b/dlm_controld/main.c
@@ -8,6 +8,7 @@
#define EXTERN
#include "dlm_daemon.h"
+#include <ctype.h>
#include <pthread.h>
#include <linux/netlink.h>
#include <linux/genetlink.h>
@@ -443,6 +444,46 @@ static void query_dump_debug(int fd)
send(fd, copy_buf, len, MSG_NOSIGNAL);
}
+static void copy_options(char *buf, int *len)
+{
+ struct dlm_option *o;
+ char tmp[256];
+ int i, ret, pos = 0;
+
+ for (i = 0; i < dlm_options_max; i++) {
+ o = &dlm_options[i];
+
+ memset(tmp, 0, sizeof(tmp));
+
+ if (o->req_arg == req_arg_str)
+ snprintf(tmp, 255, "%s=%s\n", o->name, o->use_str);
+ else
+ snprintf(tmp, 255, "%s=%d\n", o->name, o->use_int);
+
+ if (pos + strlen(tmp) >= LOG_DUMP_SIZE)
+ break;
+
+ ret = sprintf(buf + pos, "%s", tmp);
+ pos += ret;
+ }
+
+ *len = pos;
+}
+
+static void query_dump_config(int fd)
+{
+ struct dlmc_header h;
+ int len = 0;
+
+ copy_options(copy_buf, &len);
+
+ init_header(&h, DLMC_CMD_DUMP_CONFIG, NULL, 0, len);
+ send(fd, &h, sizeof(h), MSG_NOSIGNAL);
+
+ if (len)
+ send(fd, copy_buf, len, MSG_NOSIGNAL);
+}
+
static void query_dump_log_plock(int fd)
{
struct dlmc_header h;
@@ -658,7 +699,7 @@ static void process_connection(int ci)
break;
case DLMC_CMD_FS_REGISTER:
- if (cfgd_enable_fscontrol) {
+ if (opt(enable_fscontrol_ind)) {
rv = fs_register_add(h.name);
ls = find_ls(h.name);
if (ls)
@@ -805,6 +846,9 @@ static void *process_queries(void *arg)
case DLMC_CMD_DUMP_DEBUG:
query_dump_debug(f);
break;
+ case DLMC_CMD_DUMP_CONFIG:
+ query_dump_config(f);
+ break;
case DLMC_CMD_DUMP_LOG_PLOCK:
query_dump_log_plock(f);
break;
@@ -951,7 +995,7 @@ static void loop(void)
goto out;
#if 0
- if (cfgd_enable_deadlk) {
+ if (opt(enable_deadlk_ind)) {
rv = setup_netlink();
if (rv < 0)
goto out;
@@ -1114,174 +1158,367 @@ static void unlink_lockfile(int fd, const char *dir, const char *name)
close(fd);
}
+static const char *req_arg_s(int a)
+{
+ switch (a) {
+ case no_arg:
+ return "";
+ case req_arg_bool:
+ return "0|1";
+ case req_arg_int:
+ return "<int>";
+ case req_arg_str:
+ return "<str>";
+ default:
+ return "<arg>";
+ }
+}
+
static void print_usage(void)
{
+ struct dlm_option *o;
+ int i;
+
printf("Usage:\n");
printf("\n");
printf("dlm_controld [options]\n");
printf("\n");
printf("Options:\n");
printf("\n");
- printf(" -D Enable debugging to stderr and don't fork\n");
- printf(" -L Enable debugging to log file\n");
- printf(" -K Enable kernel dlm debugging messages\n");
- printf(" -r <num> dlm kernel lowcomms protocol, 0 tcp, 1 sctp, 2 detect\n");
- printf(" 2 selects tcp if corosync rrp_mode is \"none\", otherwise sctp\n");
- printf(" Default is 2\n");
-
- printf(" -j <sec> Seconds to delay fencing after cluster join\n");
- printf(" Default is %d\n", DEFAULT_POST_JOIN_DELAY);
- printf(" -f <num> Enable (1) or disable (0) fencing\n");
- printf(" Default is %d\n", DEFAULT_ENABLE_FENCING);
- printf(" -s <num> Enable (1) or disable (0) startup fencing\n");
- printf(" Default is %d\n", DEFAULT_ENABLE_STARTUP_FENCING);
- printf(" -q <num> Enable (1) or disable (0) quorum wait before fencing\n");
- printf(" Default is %d\n", DEFAULT_ENABLE_QUORUM_FENCING);
-
- printf(" -p <num> Enable (1) or disable (0) plock code for cluster fs\n");
- printf(" Default is %d\n", DEFAULT_ENABLE_PLOCK);
- printf(" -P Enable plock debugging\n");
- printf(" -l <limit> Limit the rate of plock operations\n");
- printf(" Default is %d, set to 0 for no limit\n", DEFAULT_PLOCK_RATE_LIMIT);
- printf(" -o <n> Enable (1) or disable (0) plock ownership\n");
- printf(" Default is %d\n", DEFAULT_PLOCK_OWNERSHIP);
- printf(" -t <ms> plock ownership drop resources time (milliseconds)\n");
- printf(" Default is %u\n", DEFAULT_DROP_RESOURCES_TIME);
- printf(" -c <num> plock ownership drop resources count\n");
- printf(" Default is %u\n", DEFAULT_DROP_RESOURCES_COUNT);
- printf(" -a <ms> plock ownership drop resources age (milliseconds)\n");
- printf(" Default is %u\n", DEFAULT_DROP_RESOURCES_AGE);
-
- printf(" -h Print this help, then exit\n");
- printf(" -V Print program version information, then exit\n");
-}
-#define OPTION_STRING "LDKf:q:p:Pl:o:t:c:a:hVr:s:e:j:"
+ for (i = 0; i < dlm_options_max; i++) {
+ o = &dlm_options[i];
-static void read_arguments(int argc, char **argv)
-{
- int cont = 1;
- int optchar;
+ /* don't advertise options with no description */
+ if (!strlen(o->desc))
+ continue;
- while (cont) {
- optchar = getopt(argc, argv, OPTION_STRING);
+ printf(" --%s", o->name);
- switch (optchar) {
- case 'D':
- daemon_debug_opt = 1;
- break;
+ if (o->letter) {
+ printf(" | -%c", o->letter);
+ if (o->req_arg)
+ printf(" %s", req_arg_s(o->req_arg));
+ } else {
+ if (o->req_arg)
+ printf(" %s", req_arg_s(o->req_arg));
+ }
- case 'L':
- optd_debug_logfile = 1;
- cfgd_debug_logfile = 1;
- break;
+ printf("\n");
- case 'K':
- optk_debug = 1;
- cfgk_debug = 1;
- break;
+ printf(" %s", o->desc);
- case 'r':
- optk_protocol = 1;
- cfgk_protocol = atoi(optarg);
- break;
+ if (o->req_arg == req_arg_str)
+ printf(" [%s]\n", o->default_str ? o->default_str : "");
+ else if (o->req_arg == req_arg_int)
+ printf(" [%d]\n", o->default_int);
+ else if (o->req_arg == req_arg_bool)
+ printf(" [%d]\n", o->default_int);
+ else
+ printf("\n");
+
+ printf("\n");
+ }
+}
- /* fencing options */
+static void set_opt_default(int ind, const char *name, char letter, int arg_type,
+ int default_int, const char *default_str, const char *desc)
+{
+ dlm_options[ind].name = name;
+ dlm_options[ind].letter = letter;
+ dlm_options[ind].req_arg = arg_type;
+ dlm_options[ind].desc = desc;
+ dlm_options[ind].default_int = default_int;
+ dlm_options[ind].default_str = default_str;
+ dlm_options[ind].use_int = default_int;
+ dlm_options[ind].use_str = (char *)default_str;
+}
- case 'j':
- optd_post_join_delay = 1;
- cfgd_post_join_delay = atoi(optarg);
- break;
+static void set_opt_defaults(void)
+{
+ set_opt_default(daemon_debug_ind,
+ "daemon_debug", 'D', no_arg,
+ 0, NULL,
+ "enable debugging to stderr and don't fork");
+
+ set_opt_default(log_debug_ind,
+ "log_debug", 'K', no_arg,
+ 0, NULL,
+ "enable kernel dlm debugging messages");
+
+ set_opt_default(timewarn_ind,
+ "timewarn", '\0', req_arg_int,
+ 0, NULL,
+ ""); /* do not advertise */
+
+ set_opt_default(protocol_ind,
+ "protocol", 'r', req_arg_str,
+ -1, "detect",
+ "dlm kernel lowcomms protocol: tcp, sctp, detect");
+
+ set_opt_default(debug_logfile_ind,
+ "debug_logfile", 'L', no_arg,
+ 0, NULL,
+ "write debugging to log file");
+
+ set_opt_default(enable_fscontrol_ind,
+ "enable_fscontrol", '\0', req_arg_bool,
+ 0, NULL,
+ "enable/disable recovery coordination with fs controld");
+
+ set_opt_default(enable_plock_ind,
+ "enable_plock", 'p', req_arg_bool,
+ 1, NULL,
+ "enable/disable posix lock support for cluster fs");
+
+ set_opt_default(plock_debug_ind,
+ "plock_debug", 'P', no_arg,
+ 0, NULL,
+ "enable plock debugging");
+
+ set_opt_default(plock_rate_limit_ind,
+ "plock_rate_limit", 'l', req_arg_int,
+ 0, NULL,
+ "limit rate of plock operations (0 for none)");
+
+ set_opt_default(plock_ownership_ind,
+ "plock_ownership", 'o', req_arg_bool,
+ 0, NULL,
+ "enable/disable plock ownership");
+
+ set_opt_default(drop_resources_time_ind,
+ "drop_resources_time", 't', req_arg_int,
+ 10000, NULL,
+ "plock ownership drop resources time (milliseconds)");
+
+ set_opt_default(drop_resources_count_ind,
+ "drop_resources_count", 'c', req_arg_int,
+ 10, NULL,
+ "plock ownership drop resources count");
+
+ set_opt_default(drop_resources_age_ind,
+ "drop_resources_age", 'a', req_arg_int,
+ 10000, NULL,
+ "plock ownership drop resources age (milliseconds)");
+
+ set_opt_default(post_join_delay_ind,
+ "post_join_delay", 'j', req_arg_int,
+ 30, NULL,
+ "seconds to delay fencing after cluster join");
+
+ set_opt_default(enable_fencing_ind,
+ "enable_fencing", 'f', req_arg_bool,
+ 1, NULL,
+ "enable/disable fencing");
+
+ set_opt_default(enable_concurrent_fencing_ind,
+ "enable_concurrent_fencing", '\0', req_arg_bool,
+ 0, NULL,
+ "enable/disable concurrent fencing");
+
+ set_opt_default(enable_startup_fencing_ind,
+ "enable_startup_fencing", 's', req_arg_bool,
+ 1, NULL,
+ "enable/disable startup fencing");
+
+ set_opt_default(enable_quorum_fencing_ind,
+ "enable_quorum_fencing", 'q', req_arg_bool,
+ 1, NULL,
+ "enable/disable quorum requirement for fencing");
+
+ set_opt_default(enable_quorum_lockspace_ind,
+ "enable_quorum_lockspace", '\0', req_arg_bool,
+ 1, NULL,
+ "enable/disable quorum requirement for lockspace operations");
+
+ set_opt_default(fence_all_ind,
+ "fence_all", '\0', req_arg_str,
+ -1, "dlm_stonith",
+ "fence all nodes with this agent");
+
+ set_opt_default(unfence_all_ind,
+ "unfence_all", '\0', req_arg_str,
+ -1, NULL,
+ "unfence all nodes with this agent");
+
+ set_opt_default(help_ind,
+ "help", 'h', no_arg,
+ -1, NULL,
+ "print this help, then exit");
+
+ set_opt_default(version_ind,
+ "version", 'V', no_arg,
+ -1, NULL,
+ "Print program version information, then exit");
+}
- case 'f':
- optd_enable_fencing = 1;
- cfgd_enable_fencing = atoi(optarg);
- break;
+static int get_ind_name(char *s)
+{
+ char name[PATH_MAX];
+ char *p = s;
+ int i;
- case 's':
- optd_enable_startup_fencing = 1;
- cfgd_enable_startup_fencing = atoi(optarg);
+ memset(name, 0, sizeof(name));
- case 'q':
- optd_enable_quorum_fencing = 1;
- cfgd_enable_quorum_fencing = atoi(optarg);
+ for (i = 0; i < strlen(s); i++) {
+ if (*p == '=')
break;
-
- case 'e':
- optd_fence_all_agent = 1;
- strcpy(fence_all_agent, optarg);
+ if (*p == ' ')
break;
+ name[i] = *p;
+ p++;
+ }
+
+ for (i = 0; i < dlm_options_max; i++) {
+ if (!strcmp(dlm_options[i].name, name))
+ return i;
+ }
+ return -1;
+}
+static int get_ind_letter(char c)
+{
+ int i;
- /* plock options */
+ for (i = 0; i < dlm_options_max; i++) {
+ if (dlm_options[i].letter == c)
+ return i;
+ }
+ return -1;
+}
- case 'p':
- optd_enable_plock = 1;
- cfgd_enable_plock = atoi(optarg);
- break;
+struct dlm_option *get_dlm_option(char *name)
+{
+ int i;
+ i = get_ind_name(name);
+ if (i < 0)
+ return NULL;
+ return &dlm_options[i];
+}
- case 'P':
- optd_plock_debug = 1;
- cfgd_plock_debug = 1;
- break;
+static void set_opt_cli(int argc, char **argv)
+{
+ struct dlm_option *o;
+ char *arg1 = argv[1];
+ char *p, *arg_str;
+ char bundled_letters[8];
+ int b, blc = 0, blc_max = 8;
+ int debug_options = 0;
+ int i, ind;
+
+ if (argc < 2 || !strcmp(arg1, "help") || !strcmp(arg1, "--help") || !strcmp(arg1, "-h")) {
+ print_usage();
+ exit(EXIT_SUCCESS);
+ }
- case 'l':
- optd_plock_rate_limit = 1;
- cfgd_plock_rate_limit = atoi(optarg);
- break;
+ if (!strcmp(arg1, "version") || !strcmp(arg1, "--version") || !strcmp(arg1, "-V")) {
+ printf("dlm_controld %s (built %s %s)\n",
+ RELEASE_VERSION, __DATE__, __TIME__);
+ printf("%s\n", REDHAT_COPYRIGHT);
+ exit(EXIT_SUCCESS);
+ }
- case 'o':
- optd_plock_ownership = 1;
- cfgd_plock_ownership = atoi(optarg);
- break;
+ for (i = 1; i < argc; ) {
+ p = argv[i++];
- case 't':
- optd_drop_resources_time = 1;
- cfgd_drop_resources_time = atoi(optarg);
- break;
+ if (!strcmp(p, "--debug_options")) {
+ debug_options = 1;
+ continue;
+ }
- case 'c':
- optd_drop_resources_count = 1;
- cfgd_drop_resources_count = atoi(optarg);
- break;
+ if (p[0] == '-' && p[1] == '-')
+ ind = get_ind_name(p + 2);
+ else if (p[0] == '-')
+ ind = get_ind_letter(p[1]);
+ else {
+ fprintf(stderr, "unknown option arg %s\n", p);
+ exit(EXIT_FAILURE);
+ }
- case 'a':
- optd_drop_resources_age = 1;
- cfgd_drop_resources_age = atoi(optarg);
- break;
+ if (ind < 0) {
+ fprintf(stderr, "unknown option %s\n", p);
+ exit(EXIT_FAILURE);
+ }
- case 'h':
- print_usage();
- exit(EXIT_SUCCESS);
- break;
+ o = &dlm_options[ind];
+ o->cli_set++;
+
+ if (!o->req_arg) {
+ /* "-x" has same effect as "-x 1" */
+ o->cli_int = 1;
+ o->use_int = 1;
+
+ /* save bundled, arg-less, single letters, e.g. -DKP */
+ if ((p[0] == '-') && isalpha(p[1]) && (strlen(p) > 2)) {
+ for (b = 2; b < strlen(p) && blc < blc_max; b++) {
+ if (!isalpha(p[b]))
+ break;
+ bundled_letters[blc++] = p[b];
+ }
+ }
+ continue;
+ }
- case 'V':
- printf("dlm_controld %s (built %s %s)\n",
- RELEASE_VERSION, __DATE__, __TIME__);
- printf("%s\n", REDHAT_COPYRIGHT);
- exit(EXIT_SUCCESS);
- break;
+ arg_str = NULL;
+
+ if (strstr(p, "=")) {
+ /* arg starts after = for name or letter */
+ arg_str = strstr(p, "=") + 1;
+
+ } else if (strlen(p) > 2 && isalpha(p[1]) && isdigit(p[2])) {
+ /* arg with no space between letter and digits */
+ arg_str = p + 2;
+
+ } else {
+ /* space separates arg from name or letter */
+ if (i >= argc) {
+ fprintf(stderr, "option %s no arg", p);
+ exit(EXIT_FAILURE);
+ }
+ arg_str = argv[i++];
+ }
- case ':':
- case '?':
- fprintf(stderr, "Please use '-h' for usage.\n");
+ if (!arg_str || arg_str[0] == '-' || arg_str[0] == '\0') {
+ fprintf(stderr, "option %s requires arg", p);
exit(EXIT_FAILURE);
- break;
+ }
- case EOF:
- cont = 0;
- break;
+ if (o->req_arg == req_arg_str) {
+ o->cli_str = strdup(arg_str);
+ o->use_str = o->cli_str;
+ } else if (o->req_arg == req_arg_int) {
+ o->cli_int = atoi(arg_str);
+ o->use_int = o->cli_int;
+ } else if (o->req_arg == req_arg_bool) {
+ o->cli_int = atoi(arg_str) ? 1 : 0;
+ o->use_int = o->cli_int;
+ }
+ }
- default:
- fprintf(stderr, "unknown option: %c\n", optchar);
+ /* process bundled letters saved above */
+
+ for (i = 0; i < blc; i++) {
+ ind = get_ind_letter(bundled_letters[i]);
+ if (ind < 0) {
+ fprintf(stderr, "unknown option char %c\n", bundled_letters[i]);
exit(EXIT_FAILURE);
- break;
- };
+ }
+ o = &dlm_options[ind];
+ o->cli_set++;
+ o->cli_int = 1;
+ o->use_int = 1;
+ }
+
+ if (debug_options && opt(daemon_debug_ind)) {
+ for (i = 0; i < dlm_options_max; i++) {
+ o = &dlm_options[i];
+ printf("%-25s cli_set %d cli_int %d cli_str %s use_int %d use_str %s\n",
+ o->name, o->cli_set, o->cli_int, o->cli_str, o->use_int, o->use_str);
+ }
}
if (getenv("DLM_CONTROLD_DEBUG")) {
- optd_debug_logfile = 1;
- cfgd_debug_logfile = 1;
+ dlm_options[daemon_debug_ind].use_int = 1;
}
}
@@ -1308,39 +1545,19 @@ int main(int argc, char **argv)
struct sigaction act;
int fd, rv;
- cfgk_debug = -1;
- cfgk_timewarn = -1;
- cfgk_protocol = PROTO_DETECT;
- cfgd_debug_logfile = DEFAULT_DEBUG_LOGFILE;
- cfgd_enable_fscontrol = DEFAULT_ENABLE_FSCONTROL;
-
- cfgd_enable_plock = DEFAULT_ENABLE_PLOCK;
- cfgd_plock_debug = DEFAULT_PLOCK_DEBUG;
- cfgd_plock_rate_limit = DEFAULT_PLOCK_RATE_LIMIT;
- cfgd_plock_ownership = DEFAULT_PLOCK_OWNERSHIP;
- cfgd_drop_resources_time = DEFAULT_DROP_RESOURCES_TIME;
- cfgd_drop_resources_count = DEFAULT_DROP_RESOURCES_COUNT;
- cfgd_drop_resources_age = DEFAULT_DROP_RESOURCES_AGE;
-
- cfgd_post_join_delay = DEFAULT_POST_JOIN_DELAY;
- cfgd_enable_fencing = DEFAULT_ENABLE_FENCING;
- cfgd_enable_startup_fencing = DEFAULT_ENABLE_STARTUP_FENCING;
- cfgd_enable_quorum_fencing = DEFAULT_ENABLE_QUORUM_FENCING;
- cfgd_enable_quorum_lockspace= DEFAULT_ENABLE_QUORUM_LOCKSPACE;
-
- strcpy(fence_all_agent, DEFAULT_FENCE_ALL_AGENT);
+ set_opt_defaults();
+
memset(&fence_all_device, 0, sizeof(struct fence_device));
strcpy(fence_all_device.name, "fence_all");
- strcpy(fence_all_device.agent, fence_all_agent);
+ strcpy(fence_all_device.agent, dlm_options[fence_all_ind].default_str);
INIT_LIST_HEAD(&lockspaces);
INIT_LIST_HEAD(&fs_register_list);
-
init_daemon();
- read_arguments(argc, argv);
+ set_opt_cli(argc, argv);
- if (!daemon_debug_opt) {
+ if (!opt(daemon_debug_ind)) {
if (daemon(0, 0) < 0) {
perror("daemon error");
exit(EXIT_FAILURE);
diff --git a/dlm_controld/member.c b/dlm_controld/member.c
index 7222b9e..d25ad42 100644
--- a/dlm_controld/member.c
+++ b/dlm_controld/member.c
@@ -356,7 +356,7 @@ int setup_node_config(void)
log_debug("node_config %d", nodeid);
- if (cfgd_enable_fencing && cfgd_enable_startup_fencing)
+ if (opt(enable_fencing_ind) && opt(enable_startup_fencing_ind))
add_startup_node(nodeid);
}
diff --git a/dlm_controld/plock.c b/dlm_controld/plock.c
index ed98e9c..72bf6ea 100644
--- a/dlm_controld/plock.c
+++ b/dlm_controld/plock.c
@@ -293,7 +293,7 @@ static int find_resource(struct lockspace *ls, uint64_t number, int create,
INIT_LIST_HEAD(&r->waiters);
INIT_LIST_HEAD(&r->pending);
- if (cfgd_plock_ownership)
+ if (opt(plock_ownership_ind))
r->owner = -1;
else
r->owner = 0;
@@ -310,7 +310,7 @@ static int find_resource(struct lockspace *ls, uint64_t number, int create,
static void put_resource(struct lockspace *ls, struct resource *r)
{
/* with ownership, resources are only freed via drop messages */
- if (cfgd_plock_ownership)
+ if (opt(plock_ownership_ind))
return;
if (list_empty(&r->locks) && list_empty(&r->waiters)) {
@@ -877,11 +877,11 @@ static void _receive_plock(struct lockspace *ls, struct dlm_header *hd, int len)
return;
}
- create = !cfgd_plock_ownership;
+ create = !opt(plock_ownership_ind);
rv = find_resource(ls, info.number, create, &r);
- if (rv && cfgd_plock_ownership) {
+ if (rv && opt(plock_ownership_ind)) {
/* There must have been a race with a drop, so we need to
ignore this plock op which will be resent. If we're the one
who sent the plock, we need to send_own() and put it on the
@@ -1420,7 +1420,7 @@ static int drop_resources(struct lockspace *ls)
struct timeval now;
int count = 0;
- if (!cfgd_plock_ownership)
+ if (!opt(plock_ownership_ind))
return 0;
if (list_empty(&ls->plock_resources))
@@ -1429,7 +1429,7 @@ static int drop_resources(struct lockspace *ls)
gettimeofday(&now, NULL);
if (time_diff_ms(&ls->drop_resources_last, &now) <
- cfgd_drop_resources_time)
+ opt(drop_resources_time_ind))
return 1;
ls->drop_resources_last = now;
@@ -1437,12 +1437,12 @@ static int drop_resources(struct lockspace *ls)
/* try to drop the oldest, unused resources */
list_for_each_entry_reverse(r, &ls->plock_resources, list) {
- if (count >= cfgd_drop_resources_count)
+ if (count >= opt(drop_resources_count_ind))
break;
if (r->owner && r->owner != our_nodeid)
continue;
if (time_diff_ms(&r->last_access, &now) <
- cfgd_drop_resources_age)
+ opt(drop_resources_age_ind))
continue;
if (list_empty(&r->locks) && list_empty(&r->waiters)) {
@@ -1478,18 +1478,18 @@ int limit_plocks(void)
{
struct timeval now;
- if (!cfgd_plock_rate_limit || !plock_read_count)
+ if (!opt(plock_rate_limit_ind) || !plock_read_count)
return 0;
gettimeofday(&now, NULL);
/* Every time a plock op is read from the kernel, we increment
- plock_read_count. After every cfgd_plock_rate_limit (N) reads,
+ plock_read_count. After every plock_rate_limit (N) reads,
we check the time it's taken to do those N; if the time is less than
a second, then we delay reading any more until a second is up.
This way we read a max of N ops from the kernel every second. */
- if (!(plock_read_count % cfgd_plock_rate_limit)) {
+ if (!(plock_read_count % opt(plock_rate_limit_ind))) {
if (time_diff_ms(&plock_rate_last, &now) < 1000) {
plock_rate_delays++;
return 2;
@@ -1529,7 +1529,7 @@ void process_plocks(int ci)
/* kernel doesn't set the nodeid field */
info.nodeid = our_nodeid;
- if (!cfgd_enable_plock) {
+ if (!opt(enable_plock_ind)) {
rv = -ENOSYS;
goto fail;
}
@@ -1586,7 +1586,7 @@ void process_plocks(int ci)
save_pending_plock(ls, r, &info);
}
- if (cfgd_plock_ownership && !list_empty(&ls->plock_resources))
+ if (opt(plock_ownership_ind) && !list_empty(&ls->plock_resources))
poll_drop_plock = 1;
return;
@@ -1670,7 +1670,7 @@ static int pack_send_buf(struct lockspace *ls, struct resource *r, int owner,
}
/* plocks not replicated for owned resources */
- if (cfgd_plock_ownership && (owner == our_nodeid))
+ if (opt(plock_ownership_ind) && (owner == our_nodeid))
goto done;
len = sizeof(struct dlm_header) + sizeof(struct resource_data);
@@ -1769,7 +1769,7 @@ void send_all_plocks_data(struct lockspace *ls, uint32_t seq, uint32_t *plocks_d
int owner, count, len, full;
uint32_t send_count = 0;
- if (!cfgd_enable_plock || ls->disable_plock)
+ if (!opt(enable_plock_ind) || ls->disable_plock)
return;
log_dlock(ls, "send_all_plocks_data %d:%u", our_nodeid, seq);
@@ -1785,7 +1785,7 @@ void send_all_plocks_data(struct lockspace *ls, uint32_t seq, uint32_t *plocks_d
(there should be no SYNCING plocks) */
list_for_each_entry(r, &ls->plock_resources, list) {
- if (!cfgd_plock_ownership)
+ if (!opt(plock_ownership_ind))
owner = 0;
else if (r->owner == -1)
continue;
@@ -1861,7 +1861,7 @@ void receive_plocks_data(struct lockspace *ls, struct dlm_header *hd, int len)
int owner;
int i;
- if (!cfgd_enable_plock || ls->disable_plock)
+ if (!opt(enable_plock_ind) || ls->disable_plock)
return;
if (!ls->need_plocks)
@@ -1907,7 +1907,7 @@ void receive_plocks_data(struct lockspace *ls, struct dlm_header *hd, int len)
INIT_LIST_HEAD(&r->waiters);
INIT_LIST_HEAD(&r->pending);
- if (!cfgd_plock_ownership) {
+ if (!opt(plock_ownership_ind)) {
if (owner) {
log_elock(ls, "recv_plocks_data %d:%u n %llu bad owner %d",
hd->nodeid, hd->msgdata, (unsigned long long)num,
@@ -1991,7 +1991,7 @@ void clear_plocks_data(struct lockspace *ls)
struct resource *r, *r2;
uint32_t count = 0;
- if (!cfgd_enable_plock || ls->disable_plock)
+ if (!opt(enable_plock_ind) || ls->disable_plock)
return;
list_for_each_entry_safe(r, r2, &ls->plock_resources, list) {
@@ -2019,7 +2019,7 @@ void purge_plocks(struct lockspace *ls, int nodeid, int unmount)
struct resource *r, *r2;
int purged = 0;
- if (!cfgd_enable_plock || ls->disable_plock)
+ if (!opt(enable_plock_ind) || ls->disable_plock)
return;
list_for_each_entry_safe(r, r2, &ls->plock_resources, list) {
@@ -2053,7 +2053,7 @@ void purge_plocks(struct lockspace *ls, int nodeid, int unmount)
if (!list_empty(&r->waiters))
do_waiters(ls, r);
- if (!cfgd_plock_ownership &&
+ if (!opt(plock_ownership_ind) &&
list_empty(&r->locks) && list_empty(&r->waiters)) {
rb_del_plock_resource(ls, r);
list_del(&r->list);
diff --git a/dlm_tool/main.c b/dlm_tool/main.c
index 5df5c59..1fb6b3e 100644
--- a/dlm_tool/main.c
+++ b/dlm_tool/main.c
@@ -40,6 +40,7 @@
#define OP_LOG_PLOCK 10
#define OP_FENCE_ACK 11
#define OP_STATUS 12
+#define OP_DUMP_CONFIG 13
static char *prog_name;
static char *lsname;
@@ -183,7 +184,7 @@ static void print_usage(void)
printf("Usage:\n");
printf("\n");
printf("dlm_tool [options] [join | leave | lockdump | lockdebug |\n"
- " ls | dump | log_plock | plocks |\n"
+ " ls | dump | dump_config | log_plock | plocks |\n"
" fence_ack]\n");
printf("\n");
printf("Options:\n");
@@ -340,6 +341,12 @@ static void decode_arguments(int argc, char **argv)
opt_ind = optind + 1;
need_lsname = 0;
break;
+ } else if (!strncmp(argv[optind], "dump_config", 11) &&
+ (strlen(argv[optind]) == 11)) {
+ operation = OP_DUMP_CONFIG;
+ opt_ind = optind + 1;
+ need_lsname = 0;
+ break;
} else if (!strncmp(argv[optind], "plocks", 6) &&
(strlen(argv[optind]) == 6)) {
operation = OP_PLOCKS;
@@ -1273,13 +1280,16 @@ static void do_plocks(char *name)
do_write(STDOUT_FILENO, buf, strlen(buf));
}
-static void do_dump(void)
+static void do_dump(int op)
{
char buf[DLMC_DUMP_SIZE];
memset(buf, 0, sizeof(buf));
- dlmc_dump_debug(buf);
+ if (op == OP_DUMP)
+ dlmc_dump_debug(buf);
+ else if (op == OP_DUMP_CONFIG)
+ dlmc_dump_config(buf);
buf[DLMC_DUMP_SIZE-1] = '\0';
@@ -1334,7 +1344,11 @@ int main(int argc, char **argv)
break;
case OP_DUMP:
- do_dump();
+ do_dump(operation);
+ break;
+
+ case OP_DUMP_CONFIG:
+ do_dump(operation);
break;
case OP_LOG_PLOCK:
12 years, 1 month
- 1
- 0
dlm: master - dlm_controld: waiting for fence result
by David Teigland
Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=...
Commit: b8d1a03c2d23a42761216ac33dae6e42f87ec782
Parent: 82f3290001c891e833e0b676bd4a07181c8081f9
Author: David Teigland <teigland(a)redhat.com>
AuthorDate: Tue Apr 3 12:34:06 2012 -0500
Committer: David Teigland <teigland(a)redhat.com>
CommitterDate: Tue Apr 3 12:34:06 2012 -0500
dlm_controld: waiting for fence result
Add an extra state to address the short period between
our own send_fence_result and receive_fence_result.
There's nothing to prevent us from doing fence_request
again during this time.
Signed-off-by: David Teigland <teigland(a)redhat.com>
---
dlm_controld/daemon_cpg.c | 51 +++++++++++++++++++++++++++++++-------------
1 files changed, 36 insertions(+), 15 deletions(-)
diff --git a/dlm_controld/daemon_cpg.c b/dlm_controld/daemon_cpg.c
index 9f48da0..a68de80 100644
--- a/dlm_controld/daemon_cpg.c
+++ b/dlm_controld/daemon_cpg.c
@@ -64,6 +64,7 @@ struct node_daemon {
int delay_fencing;
int fence_pid;
int fence_pid_wait;
+ int fence_result_wait;
int fence_actor_done; /* for status/debug */
int fence_actor_last; /* for status/debug */
int fence_actors[MAX_NODES];
@@ -625,7 +626,7 @@ static void fence_pid_cancel(int nodeid, int pid)
{
int rv, result = 0;
- log_debug("fence_pid_cancel nodeid %d pid %d", nodeid, pid);
+ log_debug("fence_pid_cancel nodeid %d pid %d sigkill", nodeid, pid);
kill(pid, SIGKILL);
usleep(500000);
@@ -724,11 +725,6 @@ static void fence_pid_cancel(int nodeid, int pid)
* later same as case B above
*/
-/*
- * TODO: limit to one agent running at once, in case both
- * instances need to log into the same switch, for example.
- */
-
static void daemon_fence_work(void)
{
struct node_daemon *node, *safe;
@@ -806,6 +802,7 @@ static void daemon_fence_work(void)
node->fence_actor_done = 0;
node->fence_pid_wait = 0;
node->fence_pid = 0;
+ node->fence_result_wait = 0;
node->fence_config.pos = 0;
node->left_reason = REASON_STARTUP_FENCING;
node->fail_monotime = cluster_joined_monotime - 1;
@@ -826,6 +823,11 @@ static void daemon_fence_work(void)
if (node->fence_pid_wait)
continue;
+ if (node->fence_result_wait) {
+ log_debug("fence request %d result_wait", node->nodeid);
+ continue;
+ }
+
if (is_clean_daemon_member(node->nodeid)) {
/*
* node has rejoined in clean state
@@ -887,9 +889,8 @@ static void daemon_fence_work(void)
node->left_reason,
&pid);
if (rv < 0) {
- /* FIXME: keep ourself from retrying between here
- * and receiving this message */
send_fence_result(node->nodeid, rv, 0, time(NULL));
+ node->fence_result_wait = 1;
continue;
}
@@ -909,12 +910,13 @@ static void daemon_fence_work(void)
if (node->delay_fencing)
continue;
+ if (node->fence_result_wait) {
+ log_debug("fence wait %d result_wait", node->nodeid);
+ continue;
+ }
+
if (!node->fence_pid_wait) {
- /*
- * another node is the actor, or we were actor,
- * and are between our own send_fence_result()
- * and receive_fence_result()
- */
+ /* another node is the actor */
log_debug("fence wait %d for done", node->nodeid);
continue;
}
@@ -978,15 +980,19 @@ static void daemon_fence_work(void)
parallel, set it to run next, otherwise success */
rv = fence_config_next_parallel(&node->fence_config);
- if (rv < 0)
+ if (rv < 0) {
send_fence_result(nodeid, 0, 0, time(NULL));
+ node->fence_result_wait = 1;
+ }
} else {
/* agent exit 1, if there's another agent to run at
next priority, set it to run next, otherwise fail */
rv = fence_config_next_priority(&node->fence_config);
- if (rv < 0)
+ if (rv < 0) {
send_fence_result(nodeid, result, 0, time(NULL));
+ node->fence_result_wait = 1;
+ }
}
}
@@ -1238,6 +1244,16 @@ static void receive_fence_result(struct dlm_header *hd, int len)
return;
}
+ if ((hd->nodeid == our_nodeid) && !node->fence_result_wait && (fr->result != -ECANCELED)) {
+ /* should never happen */
+ log_error("receive_fence_result %d from %d result %d no fence_result_wait",
+ fr->nodeid, hd->nodeid, fr->result);
+ /* should we ignore and return here? */
+ }
+
+ if ((hd->nodeid == our_nodeid) && (fr->result != -ECANCELED))
+ node->fence_result_wait = 0;
+
now = monotime();
log_error("fence status %d receive %d from %d walltime %llu local %llu",
@@ -1887,6 +1903,7 @@ static void confchg_cb_daemon(cpg_handle_t handle,
node->fence_actor_done = 0;
node->fence_pid_wait = 0;
node->fence_pid = 0;
+ node->fence_result_wait = 0;
node->fence_config.pos = 0;
node->left_reason = reason;
node->fail_monotime = now;
@@ -2040,6 +2057,8 @@ static int print_state_daemon_node(struct node_daemon *node, char *str)
"need_fencing=%d "
"delay_fencing=%d "
"fence_pid=%d "
+ "fence_pid_wait=%d "
+ "fence_result_wait=%d "
"fence_actor_last=%d "
"fence_actor_done=%d "
"add_time=%llu "
@@ -2054,6 +2073,8 @@ static int print_state_daemon_node(struct node_daemon *node, char *str)
node->need_fencing,
node->delay_fencing,
node->fence_pid,
+ node->fence_pid_wait,
+ node->fence_result_wait,
node->fence_actor_last,
node->fence_actor_done,
(unsigned long long)node->daemon_add_time,
12 years, 1 month
- 1
- 0
dlm: master - dlm_controld: fence config error handling
by David Teigland
Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=...
Commit: 82f3290001c891e833e0b676bd4a07181c8081f9
Parent: 113c0ac73265bab222e5c11cbd6ac17ec109cf6a
Author: David Teigland <teigland(a)redhat.com>
AuthorDate: Mon Apr 2 16:35:07 2012 -0500
Committer: David Teigland <teigland(a)redhat.com>
CommitterDate: Mon Apr 2 16:35:07 2012 -0500
dlm_controld: fence config error handling
Distinguish between errors for a config that doesn't
exist and a config with problems.
Signed-off-by: David Teigland <teigland(a)redhat.com>
---
dlm_controld/daemon_cpg.c | 13 ++++++++++---
dlm_controld/fence.c | 9 ++++++++-
dlm_controld/fence_config.c | 36 ++++++++++++++++++++++++++----------
dlm_controld/fence_config.h | 10 ++++++++++
4 files changed, 54 insertions(+), 14 deletions(-)
diff --git a/dlm_controld/daemon_cpg.c b/dlm_controld/daemon_cpg.c
index 0339b5d..9f48da0 100644
--- a/dlm_controld/daemon_cpg.c
+++ b/dlm_controld/daemon_cpg.c
@@ -434,6 +434,7 @@ static struct node_daemon *add_node_daemon(int nodeid)
{
struct node_daemon *node;
struct fence_config *fc;
+ int rv;
node = get_node_daemon(nodeid);
if (node)
@@ -462,12 +463,18 @@ static struct node_daemon *add_node_daemon(int nodeid)
/* explicit config file setting has second priority */
- fence_config_init(fc, (unsigned int)nodeid, (char *)CONF_FILE_PATH);
+ rv = fence_config_init(fc, (unsigned int)nodeid, (char *)CONF_FILE_PATH);
+ if (!rv)
+ goto out;
- /* no command line, no config file, use default, third priority */
+ /* no command line, no config file, so use default */
- if (!fc->dev[0] && fence_all_agent[0])
+ if (rv == -ENOENT) {
fc->dev[0] = &fence_all_device;
+ goto out;
+ }
+
+ log_error("fence config %d error %d", nodeid, rv);
out:
return node;
}
diff --git a/dlm_controld/fence.c b/dlm_controld/fence.c
index 633b8d4..bf5d316 100644
--- a/dlm_controld/fence.c
+++ b/dlm_controld/fence.c
@@ -188,8 +188,15 @@ int unfence_node(int nodeid)
memset(&config, 0, sizeof(config));
rv = fence_config_init(&config, nodeid, (char *)CONF_FILE_PATH);
- if (rv < 0)
+ if (rv == -ENOENT) {
+ /* file doesn't exist or doesn't contain config for nodeid */
return 0;
+ }
+ if (rv < 0) {
+ /* there's a problem with the config */
+ log_error("fence config %d error %d", nodeid, rv);
+ return rv;
+ }
memset(action, 0, sizeof(action));
snprintf(action, FENCE_CONFIG_NAME_MAX-1, "action=on\n");
diff --git a/dlm_controld/fence_config.c b/dlm_controld/fence_config.c
index 0c1782d..f886630 100644
--- a/dlm_controld/fence_config.c
+++ b/dlm_controld/fence_config.c
@@ -137,14 +137,15 @@ static int read_config_section(unsigned int nodeid, FILE *file, char *dev_line,
continue;
if (!strncmp(line, "unfence", strlen("unfence"))) {
- /* TODO: verify dev_name matches */
+ if (!strstr(line, dev_name))
+ return -EINVAL;
unfence = 1;
continue;
}
/* invalid config */
if (strncmp(line, "connect", strlen("connect")))
- return -1;
+ return -EINVAL;
/* once we've found the connect line we want, continue
scanning lines until end of section so we pick up an
@@ -158,11 +159,11 @@ static int read_config_section(unsigned int nodeid, FILE *file, char *dev_line,
rv = sscanf(line, "%s %s %[^\n]s", unused, con_name, con_args);
if (rv < 3)
- return -1;
+ return -EINVAL;
/* invalid config */
if (strncmp(dev_name, con_name, FENCE_CONFIG_NAME_MAX))
- return -1;
+ return -EINVAL;
/* skip connection for another node */
if (con_args_nodeid(con_args) != nodeid)
@@ -197,8 +198,8 @@ static int read_config_section(unsigned int nodeid, FILE *file, char *dev_line,
if (dev)
return 0;
-
- return -1;
+ else
+ return -ENOENT;
}
void fence_config_free(struct fence_config *fc)
@@ -232,7 +233,7 @@ int fence_config_init(struct fence_config *fc, unsigned int nodeid, char *path)
file = fopen(path, "r");
if (!file)
- return -1;
+ return -ENOENT;
while (fgets(line, MAX_LINE, file)) {
if (line[0] == '#')
@@ -272,18 +273,33 @@ int fence_config_init(struct fence_config *fc, unsigned int nodeid, char *path)
if (strncmp(line, "device", strlen("device")))
continue;
- /* read connect and unfence lines following a device line */
+ dev = NULL;
+ con = NULL;
+ /* read connect and unfence lines following a device line */
rv = read_config_section(nodeid, file, line, &dev, &con);
- if (rv < 0)
+
+ /* nodeid not listed in this section */
+ if (rv == -ENOENT)
continue;
+ /* an error parsing the section, may be config to free */
+ if (rv < 0) {
+ if (dev)
+ free(dev);
+ if (con)
+ free(con);
+ goto out;
+ }
+
fc->dev[pos] = dev;
fc->con[pos] = con;
pos++;
}
- if (pos)
+ if (!pos)
+ rv = -ENOENT;
+ else
rv = 0;
out:
fclose(file);
diff --git a/dlm_controld/fence_config.h b/dlm_controld/fence_config.h
index 47e7bda..b55a964 100644
--- a/dlm_controld/fence_config.h
+++ b/dlm_controld/fence_config.h
@@ -35,7 +35,17 @@ struct fence_config {
};
+/*
+ * Returns -ENOENT if path does not exist or there is no
+ * config for nodeid in the file.
+ *
+ * Returns -EXYZ if there's a problem with the config.
+ *
+ * Returns 0 if a config was found with no problems.
+ */
+
int fence_config_init(struct fence_config *fc, unsigned int nodeid, char *path);
+
void fence_config_free(struct fence_config *fc);
/*
12 years, 1 month
- 1
- 0
dlm: master - dlm_controld: unfence self at startup
by David Teigland
Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=...
Commit: 113c0ac73265bab222e5c11cbd6ac17ec109cf6a
Parent: 3d7b2195b78fa5c70e94024c777e1c178abd0282
Author: David Teigland <teigland(a)redhat.com>
AuthorDate: Mon Apr 2 15:38:04 2012 -0500
Committer: David Teigland <teigland(a)redhat.com>
CommitterDate: Mon Apr 2 15:38:04 2012 -0500
dlm_controld: unfence self at startup
Signed-off-by: David Teigland <teigland(a)redhat.com>
---
dlm_controld/dlm_daemon.h | 1 +
dlm_controld/fence.c | 70 ++++++++++++++++++++++++++++++++++++++++++-
dlm_controld/fence_config.c | 39 +++++++++++++++++++++---
dlm_controld/main.c | 19 +++++++++--
4 files changed, 119 insertions(+), 10 deletions(-)
diff --git a/dlm_controld/dlm_daemon.h b/dlm_controld/dlm_daemon.h
index f0ce9b4..4f3e551 100644
--- a/dlm_controld/dlm_daemon.h
+++ b/dlm_controld/dlm_daemon.h
@@ -402,6 +402,7 @@ int setup_node_config(void);
int fence_request(int nodeid, uint64_t fail_walltime, uint64_t fail_monotime,
struct fence_config *fc, int reason, int *pid_out);
int fence_result(int nodeid, int pid, int *result);
+int unfence_node(int nodeid);
/* netlink.c */
int setup_netlink(void);
diff --git a/dlm_controld/fence.c b/dlm_controld/fence.c
index 9807d66..633b8d4 100644
--- a/dlm_controld/fence.c
+++ b/dlm_controld/fence.c
@@ -98,7 +98,7 @@ int fence_request(int nodeid, uint64_t fail_walltime, uint64_t fail_monotime,
rv = fence_config_agent_args(fc, extra, args);
if (rv < 0) {
- log_error("fence request %d args error %d", nodeid, rv);
+ log_error("fence request %d config args error %d", nodeid, rv);
return rv;
}
@@ -176,3 +176,71 @@ int fence_result(int nodeid, int pid, int *result)
return -1;
}
+int unfence_node(int nodeid)
+{
+ struct fence_config config;
+ struct fence_device *dev;
+ char args[FENCE_CONFIG_ARGS_MAX];
+ char action[FENCE_CONFIG_NAME_MAX];
+ int rv, i, pid, status;
+ int error = 0;
+
+ memset(&config, 0, sizeof(config));
+
+ rv = fence_config_init(&config, nodeid, (char *)CONF_FILE_PATH);
+ if (rv < 0)
+ return 0;
+
+ memset(action, 0, sizeof(action));
+ snprintf(action, FENCE_CONFIG_NAME_MAX-1, "action=on\n");
+
+ for (i = 0; i < FENCE_CONFIG_DEVS_MAX; i++) {
+ dev = config.dev[i];
+ if (!dev)
+ break;
+ if (!dev->unfence)
+ continue;
+
+ config.pos = i;
+
+ memset(args, 0, sizeof(args));
+
+ rv = fence_config_agent_args(&config, action, args);
+ if (rv < 0) {
+ log_error("unfence %d config args error %d", nodeid, rv);
+ error = -1;
+ break;
+ }
+
+ rv = run_agent(dev->agent, args, &pid);
+ if (rv < 0) {
+ log_error("unfence %d %s %s run error %d", nodeid,
+ dev->name, dev->agent, rv);
+ error = -1;
+ break;
+ }
+
+ log_error("unfence %d pid %d %s %s", nodeid, pid,
+ dev->name, dev->agent);
+
+ rv = waitpid(pid, &status, 0);
+ if (rv < 0) {
+ log_error("unfence %d pid %d waitpid errno %d",
+ nodeid, pid, errno);
+ error = -1;
+ break;
+ }
+
+ if (!WIFEXITED(status) || WEXITSTATUS(status)) {
+ log_error("unfence %d pid %d error status %d",
+ nodeid, pid, status);
+ error = -1;
+ break;
+ }
+ }
+
+ fence_config_free(&config);
+
+ return error;
+}
+
diff --git a/dlm_controld/fence_config.c b/dlm_controld/fence_config.c
index 9f72e36..0c1782d 100644
--- a/dlm_controld/fence_config.c
+++ b/dlm_controld/fence_config.c
@@ -23,9 +23,14 @@ Empty new line separates the config for each fence device.
fence_all fence_foo key=val ...
-Special fence config format that applies to all nodes, allows
-no per node config parameters, and cannot be used with any
-other fence device configuration.
+Special fence config format that applies to all nodes and allows
+no per node config parameters. Multiple fence devices (parallel
+or priority) cannot be used with fence_all.
+
+fence_all fence_foo ...
+unfence_all
+
+Apply unfencing to all nodes.
-
@@ -35,6 +40,12 @@ connect <dev_name> node=<nodeid> <con_args>
General fence config format, allowing per node config
parameters.
+device <dev_name> <agent> <dev_args>
+connect <dev_name> node=<nodeid> <con_args>
+unfence <dev_name>
+
+Apply unfencing to all nodes connected to this device.
+
-
device foo fence_foo ipaddr=1.1.1.1 login=x password=y
@@ -95,7 +106,7 @@ static int read_config_section(unsigned int nodeid, FILE *file, char *dev_line,
struct fence_device **dev_out,
struct fence_connect **con_out)
{
- struct fence_device *dev;
+ struct fence_device *dev = NULL;
struct fence_connect *con;
char line[MAX_LINE];
char unused[FENCE_CONFIG_NAME_MAX];
@@ -120,6 +131,8 @@ static int read_config_section(unsigned int nodeid, FILE *file, char *dev_line,
while (fgets(line, MAX_LINE, file)) {
if (line[0] == '\n')
break;
+ if (line[0] == ' ')
+ break;
if (line[0] == '#')
continue;
@@ -133,6 +146,13 @@ static int read_config_section(unsigned int nodeid, FILE *file, char *dev_line,
if (strncmp(line, "connect", strlen("connect")))
return -1;
+ /* once we've found the connect line we want, continue
+ scanning lines until end of section so we pick up an
+ unfence line at the end */
+
+ if (dev)
+ continue;
+
memset(con_name, 0, sizeof(con_name));
memset(con_args, 0, sizeof(con_args));
@@ -170,9 +190,14 @@ static int read_config_section(unsigned int nodeid, FILE *file, char *dev_line,
*dev_out = dev;
*con_out = con;
- return 0;
}
+ if (dev && unfence)
+ dev->unfence = 1;
+
+ if (dev)
+ return 0;
+
return -1;
}
@@ -234,6 +259,10 @@ int fence_config_init(struct fence_config *fc, unsigned int nodeid, char *path)
goto out;
}
+ if (fgets(line, MAX_LINE, file) &&
+ !strncmp(line, "unfence_all", strlen("unfence_all")))
+ dev->unfence = 1;
+
fc->dev[0] = dev;
fc->pos = 0;
rv = 0;
diff --git a/dlm_controld/main.c b/dlm_controld/main.c
index 794a5c8..dcb01ec 100644
--- a/dlm_controld/main.c
+++ b/dlm_controld/main.c
@@ -898,19 +898,30 @@ static void loop(void)
if (rv > 0)
client_add(rv, process_cluster_cfg, cluster_dead);
- rv = setup_node_config();
+ rv = check_uncontrolled_lockspaces();
if (rv < 0)
goto out;
- rv = setup_cluster();
+ /*
+ * unfence needs to happen after checking for uncontrolled dlm kernel
+ * state (for which we are probably currently fenced, the state must
+ * be cleared by a reboot). unfence needs to happen before joining
+ * the daemon cpg, after which it needs to be possible for someone to
+ * fence us.
+ */
+ rv = unfence_node(our_nodeid);
if (rv < 0)
goto out;
- client_add(rv, process_cluster, cluster_dead);
- rv = check_uncontrolled_lockspaces();
+ rv = setup_node_config();
if (rv < 0)
goto out;
+ rv = setup_cluster();
+ if (rv < 0)
+ goto out;
+ client_add(rv, process_cluster, cluster_dead);
+
rv = setup_misc_devices();
if (rv < 0)
goto out;
12 years, 1 month
- 1
- 0
fence-agents: RHEL6 - fence_ipmilan: Fix typo
by Marek Grác
Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=...
Commit: d2cbc1199207c4075345c51ba159cb9d5801f014
Parent: 16efa6bcc20d5dce79559e9f82ccb658755c72fe
Author: Marek 'marx' Grac <mgrac(a)redhat.com>
AuthorDate: Mon Mar 26 15:47:25 2012 +0200
Committer: Marek 'marx' Grac <mgrac(a)redhat.com>
CommitterDate: Mon Apr 2 16:33:08 2012 +0200
fence_ipmilan: Fix typo
Resolves: rhbz#806912
---
fence/agents/ipmilan/ipmilan.c | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/fence/agents/ipmilan/ipmilan.c b/fence/agents/ipmilan/ipmilan.c
index a018d4d..ef2b582 100644
--- a/fence/agents/ipmilan/ipmilan.c
+++ b/fence/agents/ipmilan/ipmilan.c
@@ -1142,7 +1142,7 @@ main(int argc, char **argv)
if (strcasecmp(method, "onoff") &&
strcasecmp(method, "cycle")) {
- fail_exit("method, if included, muse be 'onoff', 'cycle'.");
+ fail_exit("method, if included, must be 'onoff', 'cycle'.");
}
if (!strcasecmp(method, "cycle") &&
12 years, 1 month
- 1
- 0
fence-agents: RHEL6 - fence_ipmilan: return code can be invalid when using -M cycle
by Marek Grác
Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=...
Commit: 16efa6bcc20d5dce79559e9f82ccb658755c72fe
Parent: 159a6975e6f9df039913264f436d5eb597387eea
Author: Marek 'marx' Grac <mgrac(a)redhat.com>
AuthorDate: Mon Mar 26 15:03:39 2012 +0200
Committer: Marek 'marx' Grac <mgrac(a)redhat.com>
CommitterDate: Mon Apr 2 16:32:13 2012 +0200
fence_ipmilan: return code can be invalid when using -M cycle
Resolves: rhbz#806897
---
fence/agents/ipmilan/ipmilan.c | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/fence/agents/ipmilan/ipmilan.c b/fence/agents/ipmilan/ipmilan.c
index b99e459..a018d4d 100644
--- a/fence/agents/ipmilan/ipmilan.c
+++ b/fence/agents/ipmilan/ipmilan.c
@@ -1179,6 +1179,7 @@ main(int argc, char **argv)
if (!strcasecmp(method, "cycle")) {
ret = ipmi_cycle(i);
+ translated_ret = (ret==0?ERR_OFF_SUCCESSFUL:ERR_OFF_FAIL);
} else {
/* Original onoff method */
ret = ipmi_off(i);
12 years, 1 month
- 1
- 0
fence-agents: RHEL6 - fence agents: Attribute unique should be set to 0 (false) by default
by Marek Grác
Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=...
Commit: 159a6975e6f9df039913264f436d5eb597387eea
Parent: b234ca619729b4dc68ec1d150c5052c46b703e20
Author: Marek 'marx' Grac <mgrac(a)redhat.com>
AuthorDate: Mon Mar 26 14:36:57 2012 +0200
Committer: Marek 'marx' Grac <mgrac(a)redhat.com>
CommitterDate: Mon Apr 2 16:31:56 2012 +0200
fence agents: Attribute unique should be set to 0 (false) by default
Resolves: rhbz#806883
---
fence/agents/ipmilan/ipmilan.c | 2 +-
fence/agents/lib/fencing.py.py | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/fence/agents/ipmilan/ipmilan.c b/fence/agents/ipmilan/ipmilan.c
index 63d75df..b99e459 100644
--- a/fence/agents/ipmilan/ipmilan.c
+++ b/fence/agents/ipmilan/ipmilan.c
@@ -914,7 +914,7 @@ static void print_xml_metadata(char *pname) {
printf("%s\n","<parameters>");
for (i=0;i<(sizeof(xml_parameters)/sizeof(struct xml_parameter_s));i++) {
- printf("\t<parameter name=\"%s\" unique=\"1\">\n",xml_parameters[i].name);
+ printf("\t<parameter name=\"%s\" unique=\"0\">\n",xml_parameters[i].name);
printf("\t\t<getopt mixed=\"%s\" />\n",xml_parameters[i].getopt);
if (xml_parameters[i].default_value == NULL) {
diff --git a/fence/agents/lib/fencing.py.py b/fence/agents/lib/fencing.py.py
index 02a01be..5748391 100644
--- a/fence/agents/lib/fencing.py.py
+++ b/fence/agents/lib/fencing.py.py
@@ -474,7 +474,7 @@ def metadata(avail_opt, options, docs):
print "<parameters>"
for option, value in sorted_list:
if all_opt[option].has_key("shortdesc"):
- print "\t<parameter name=\"" + option + "\" unique=\"1\" required=\"" + all_opt[option]["required"] + "\">"
+ print "\t<parameter name=\"" + option + "\" unique=\"0\" required=\"" + all_opt[option]["required"] + "\">"
default = ""
if all_opt[option].has_key("default"):
12 years, 1 month
- 1
- 0
cluster: RHEL6 - man: update fenced.8 to reflect a limitation in XML/DTD implementation
by Fabio M. Di Nitto
Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=...
Commit: 54a29913c5de797da6adb69e03b38487fef451b4
Parent: ff280bc349cb7ce973aaab1f1a3dd1b289dff711
Author: Fabio M. Di Nitto <fdinitto(a)redhat.com>
AuthorDate: Mon Apr 2 13:47:41 2012 +0200
Committer: Fabio M. Di Nitto <fdinitto(a)redhat.com>
CommitterDate: Mon Apr 2 13:49:28 2012 +0200
man: update fenced.8 to reflect a limitation in XML/DTD implementation
Resolves: rhbz#808441
Reviewed-by: Christine Caulfield <ccaulfie(a)redhat.com>
Signed-off-by: Fabio M. Di Nitto <fdinitto(a)redhat.com>
---
fence/man/fenced.8 | 5 +++++
1 files changed, 5 insertions(+), 0 deletions(-)
diff --git a/fence/man/fenced.8 b/fence/man/fenced.8
index 64f26e0..b172151 100644
--- a/fence/man/fenced.8
+++ b/fence/man/fenced.8
@@ -335,6 +335,11 @@ power never being fully disabled to the node.
</clusternode>
.fi
+.SS NOTES
+
+Due to a limitation in XML/DTD validation, the name="" value within
+the device or fencedevice section cannot start with a number.
+
.SS Hardware-specific settings
Find documentation for configuring specific devices from the device
12 years, 1 month
- 1
- 0
cluster: STABLE32 - man: update fenced.8 to reflect a limitation in XML/DTD implementation
by Fabio M. Di Nitto
Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=...
Commit: 413f459962c9bb5a9f90fd2e09b56253d5120cc6
Parent: 65a0eef71377d2de4ad94e8f20065c4467a441e9
Author: Fabio M. Di Nitto <fdinitto(a)redhat.com>
AuthorDate: Mon Apr 2 13:47:41 2012 +0200
Committer: Fabio M. Di Nitto <fdinitto(a)redhat.com>
CommitterDate: Mon Apr 2 13:47:41 2012 +0200
man: update fenced.8 to reflect a limitation in XML/DTD implementation
Resolves: rhbz#808441
Signed-off-by: Fabio M. Di Nitto <fdinitto(a)redhat.com>
---
fence/man/fenced.8 | 5 +++++
1 files changed, 5 insertions(+), 0 deletions(-)
diff --git a/fence/man/fenced.8 b/fence/man/fenced.8
index 64f26e0..b172151 100644
--- a/fence/man/fenced.8
+++ b/fence/man/fenced.8
@@ -335,6 +335,11 @@ power never being fully disabled to the node.
</clusternode>
.fi
+.SS NOTES
+
+Due to a limitation in XML/DTD validation, the name="" value within
+the device or fencedevice section cannot start with a number.
+
.SS Hardware-specific settings
Find documentation for configuring specific devices from the device
12 years, 1 month
- 1
- 0