Send Linux-ha-cvs mailing list submissions to
[email protected]
To subscribe or unsubscribe via the World Wide Web, visit
http://lists.community.tummy.com/mailman/listinfo/linux-ha-cvs
or, via email, send a message with subject or body 'help' to
[EMAIL PROTECTED]
You can reach the person managing the list at
[EMAIL PROTECTED]
When replying, please edit your Subject line so it is more specific
than "Re: Contents of Linux-ha-cvs digest..."
Today's Topics:
1. Linux-HA CVS: membership by zhenh from
([email protected])
2. Linux-HA CVS: lrm by sunjd from ([email protected])
----------------------------------------------------------------------
Message: 1
Date: Mon, 13 Mar 2006 01:14:54 -0700 (MST)
From: [email protected]
Subject: [Linux-ha-cvs] Linux-HA CVS: membership by zhenh from
To: [EMAIL PROTECTED]
Message-ID: <[EMAIL PROTECTED]>
linux-ha CVS committal
Author : zhenh
Host :
Project : linux-ha
Module : membership
Dir : linux-ha/membership/ccm
Modified Files:
ccmlib_memapi.c
Log Message:
when we recover from split brain, the born on value may reduce. so we shouldn't
use it to judge the node's presence.
===================================================================
RCS file: /home/cvs/linux-ha/linux-ha/membership/ccm/ccmlib_memapi.c,v
retrieving revision 1.41
retrieving revision 1.42
diff -u -3 -r1.41 -r1.42
--- ccmlib_memapi.c 9 Mar 2006 10:01:58 -0000 1.41
+++ ccmlib_memapi.c 13 Mar 2006 08:14:53 -0000 1.42
@@ -1,4 +1,4 @@
-/* $Id: ccmlib_memapi.c,v 1.41 2006/03/09 10:01:58 zhenh Exp $ */
+/* $Id: ccmlib_memapi.c,v 1.42 2006/03/13 08:14:53 zhenh Exp $ */
/*
* ccmlib_memapi.c: Consensus Cluster Membership API
*
@@ -214,8 +214,7 @@
{
uint i;
for ( i = 0 ; i < size ; i ++ ) {
- if(arr[i].node_id == node.node_id &&
- arr[i].node_born_on >= node.node_born_on) {
+ if(arr[i].node_id == node.node_id) {
return TRUE;
}
}
------------------------------
Message: 2
Date: Mon, 13 Mar 2006 01:35:19 -0700 (MST)
From: [email protected]
Subject: [Linux-ha-cvs] Linux-HA CVS: lrm by sunjd from
To: [EMAIL PROTECTED]
Message-ID: <[EMAIL PROTECTED]>
linux-ha CVS committal
Author : sunjd
Host :
Project : linux-ha
Module : lrm
Dir : linux-ha/lrm/lrmd
Modified Files:
lrmd.c
Log Message:
Reworkcl with new implementation for
bug756: All output from resource agents should be logged
===================================================================
RCS file: /home/cvs/linux-ha/linux-ha/lrm/lrmd/lrmd.c,v
retrieving revision 1.216
retrieving revision 1.217
diff -u -3 -r1.216 -r1.217
--- lrmd.c 10 Feb 2006 06:50:23 -0000 1.216
+++ lrmd.c 13 Mar 2006 08:35:13 -0000 1.217
@@ -1,4 +1,4 @@
-/* $Id: lrmd.c,v 1.216 2006/02/10 06:50:23 alan Exp $ */
+/* $Id: lrmd.c,v 1.217 2006/03/13 08:35:13 sunjd Exp $ */
/*
* Local Resource Manager Daemon
*
@@ -189,6 +189,7 @@
typedef struct lrmd_rsc lrmd_rsc_t;
typedef struct lrmd_op lrmd_op_t;
+typedef struct ra_pipe_op ra_pipe_op_t;
struct lrmd_rsc
@@ -213,17 +214,13 @@
pid_t client_id;
int call_id;
int exec_pid;
- char first_line_ra_stdout[80]; /* only for heartbeat RAs */
guint timeout_tag;
guint repeat_timeout_tag;
int interval;
int delay;
struct ha_msg* msg;
- /* For read the output from RAs */
- int ra_stdout_fd;
- int ra_stderr_fd;
- GFDSource * ra_stdout_gsource;
- GFDSource * ra_stderr_gsource;
+ ra_pipe_op_t * rapop;
+ char first_line_ra_stdout[80]; /* only for heartbeat RAs */
/*time stamp*/
longclock_t t_recv;
longclock_t t_addtolist;
@@ -231,6 +228,24 @@
longclock_t t_done;
};
+
+/* For read the output from RAs' execution ( its children ) */
+struct ra_pipe_op
+{
+ /* The same value of the one in corresponding lrmd_op */
+ lrmd_op_t * lrmd_op;
+ int ra_stdout_fd;
+ int ra_stderr_fd;
+ GFDSource * ra_stdout_gsource;
+ GFDSource * ra_stderr_gsource;
+ gboolean first_line_read;
+
+ /* For providing more detailed information in log */
+ char * rsc_id;
+ char * op_type;
+ char * rsc_class;
+};
+
/* Debug oriented funtions */
static gboolean debug_level_adjust(int nsig, gpointer user_data);
static void dump_data_for_debug(void);
@@ -242,10 +257,8 @@
static gboolean on_op_timeout_expired(gpointer data);
static gboolean on_repeat_op_readytorun(gpointer data);
static void on_remove_client(gpointer user_data);
-#if 0
static void destroy_pipe_ra_stderr(gpointer user_data);
static void destroy_pipe_ra_stdout(gpointer user_data);
-#endif
/* message handlers */
static int on_msg_register(lrmd_client_t* client, struct ha_msg* msg);
@@ -380,70 +393,96 @@
, lrm_objectstats.rsccount);
}
-static void
-lrmd_op_destroy(lrmd_op_t* op)
+static ra_pipe_op_t *
+ra_pipe_op_new(int child_stdout, int child_stderr, lrmd_op_t * lrmd_op)
{
+ int flag;
+ ra_pipe_op_t * rapop;
+ lrmd_rsc_t* rsc = NULL;
- CHECK_ALLOCATED(op, "op", );
- --lrm_objectstats.opcount;
+ if ( NULL == lrmd_op ) {
+ lrmd_log(LOG_WARNING
+ , "%s:%d: lrmd_op==NULL, no need to malloc ra_pipe_op"
+ , __FUNCTION__, __LINE__);
+ return NULL;
+ }
+ rapop = cl_calloc(sizeof(ra_pipe_op_t), 1);
+ if ( rapop == NULL) {
+ lrmd_log(LOG_ERR, "%s:%d out of memory."
+ , __FUNCTION__, __LINE__);
+ return NULL;
+ }
+ rapop->first_line_read = FALSE;
- if (op->exec_pid > 1) {
- return_to_orig_privs();
- if (kill(-op->exec_pid, SIGKILL) < 0 && errno != ESRCH) {
- cl_perror("Cannot kill pid %d", op->exec_pid);
+ /*
+ * No any obviouse proof of lrmd hang in pipe read yet.
+ * Bug 475 may be a duplicate of bug 499.
+ * Anyway, via test, it's proved that NOBLOCK read will
+ * obviously reduce the RA execution time (bug 553).
+ */
+ /* Let the read operation be NONBLOCK */
+ if ((flag = fcntl(child_stdout, F_GETFL)) >= 0) {
+ if (fcntl(child_stdout, F_SETFL, flag|O_NONBLOCK) < 0) {
+ cl_perror("%s::%d: fcntl", __FUNCTION__
+ , __LINE__);
}
- return_to_dropped_privs();
- return;
+ } else {
+ cl_perror("%s::%d: fcntl", __FUNCTION__, __LINE__);
}
- if ((int)op->repeat_timeout_tag > 0) {
- Gmain_timeout_remove(op->repeat_timeout_tag);
- op->repeat_timeout_tag =(guint)-1;
+ if ((flag = fcntl(child_stderr, F_GETFL)) >= 0) {
+ if (fcntl(child_stderr, F_SETFL, flag|O_NONBLOCK) < 0) {
+ cl_perror("%s::%d: fcntl", __FUNCTION__, __LINE__);
+ }
+ } else {
+ cl_perror("%s::%d: fcntl", __FUNCTION__, __LINE__);
}
- if ((int)op->timeout_tag > 0) {
- Gmain_timeout_remove(op->timeout_tag);
- op->timeout_tag = (guint)-1;
+ rapop->ra_stdout_fd = child_stdout;
+ rapop->ra_stderr_fd = child_stderr;
+ if (rapop->ra_stdout_fd <= STDERR_FILENO) {
+ lrmd_log(LOG_ERR, "%s: invalid stdout fd [%d]"
+ , __FUNCTION__, rapop->ra_stdout_fd);
+ }
+ if (rapop->ra_stderr_fd <= STDERR_FILENO) {
+ lrmd_log(LOG_ERR, "%s: invalid stderr fd [%d]"
+ , __FUNCTION__, rapop->ra_stderr_fd);
}
+
+ rapop->ra_stdout_gsource = G_main_add_fd(G_PRIORITY_HIGH
+ , child_stdout, FALSE, handle_pipe_ra_stdout
+ , rapop, destroy_pipe_ra_stdout);
+ rapop->ra_stderr_gsource = G_main_add_fd(G_PRIORITY_HIGH
+ , child_stderr, FALSE, handle_pipe_ra_stderr
+ , rapop, destroy_pipe_ra_stderr);
+
+ rapop->lrmd_op = lrmd_op;
- ha_msg_del(op->msg);
- op->msg = NULL;
- cl_free(op->rsc_id);
- op->rsc_id = NULL;
- op->exec_pid = 0;
+ rapop->op_type = cl_strdup(ha_msg_value(lrmd_op->msg, F_LRM_OP));
+ rapop->rsc_id = cl_strdup(lrmd_op->rsc_id);
+ rsc = lookup_rsc(lrmd_op->rsc_id);
+ if (rsc == NULL) {
+ lrmd_debug(LOG_WARNING
+ , "%s::%d: the rsc (id=%s) does not exist"
+ , __FUNCTION__, __LINE__, lrmd_op->rsc_id);
+ rapop->rsc_class = NULL;
+ } else {
+ rapop->rsc_class = cl_strdup(rsc->class);
+ }
+
+ return rapop;
+}
+
+static void
+ra_pipe_op_destroy(ra_pipe_op_t * op)
+{
+ CHECK_ALLOCATED(op, "ra_pipe_op", );
- /*
- * This action prohibits capturing output after the
- * process has died. FIXME!
- * Note that really fixing this would require that
- * the GSource become a semi-independent object
- * whose lifetime is potentially greater than that
- * of the operation which spawned it as described
- * by bug 756.
- * Note that if you simply remove the _del_fd()
- * call below that bad things will happen, because
- * the Gsource code refers directly to this operation
- * structure which will have been freed in
- * just a few lines below...
- */
if ( NULL != op->ra_stdout_gsource) {
G_main_del_fd(op->ra_stdout_gsource);
op->ra_stdout_gsource = NULL;
}
- /*
- * This action prohibits capturing output after the
- * process has died. FIXME!
- * Note that really fixing this would require that
- * the GSource become a semi-independent object
- * whose lifetime is potentially greater than that
- * of the operation which spawned it as described
- * by bug 756.
- * Note that if you simply remove the _del_fd()
- * call below that bad things will happen, because
- * the Gsource code refers directly to this operation
- * structure which will have been freed in just a
- * few lines below...
- */
+
if ( NULL != op->ra_stderr_gsource) {
G_main_del_fd(op->ra_stderr_gsource);
op->ra_stderr_gsource = NULL;
@@ -463,7 +502,58 @@
lrmd_log(LOG_ERR, "%s: invalid stderr fd %d"
, __FUNCTION__, op->ra_stderr_fd);
}
+ op->first_line_read = FALSE;
+
+ cl_free(op->rsc_id);
+ op->rsc_id = NULL;
+ cl_free(op->op_type);
+ op->op_type = NULL;
+ cl_free(op->rsc_class);
+ op->rsc_class = NULL;
+
+ if (op->lrmd_op != NULL) {
+ op->lrmd_op->rapop = NULL;
+ op->lrmd_op = NULL;
+ }
+
+ cl_free(op);
+}
+
+static void
+lrmd_op_destroy(lrmd_op_t* op)
+{
+
+ CHECK_ALLOCATED(op, "op", );
+ --lrm_objectstats.opcount;
+
+ if (op->exec_pid > 1) {
+ return_to_orig_privs();
+ if (kill(-op->exec_pid, SIGKILL) < 0 && errno != ESRCH) {
+ cl_perror("Cannot kill pid %d", op->exec_pid);
+ }
+ return_to_dropped_privs();
+ return;
+ }
+ if ((int)op->repeat_timeout_tag > 0) {
+ Gmain_timeout_remove(op->repeat_timeout_tag);
+ op->repeat_timeout_tag =(guint)-1;
+ }
+
+ if ((int)op->timeout_tag > 0) {
+ Gmain_timeout_remove(op->timeout_tag);
+ op->timeout_tag = (guint)-1;
+ }
+
+ ha_msg_del(op->msg);
+ op->msg = NULL;
+ cl_free(op->rsc_id);
+ op->rsc_id = NULL;
+ op->exec_pid = 0;
+ if ( op->rapop != NULL ) {
+ op->rapop->lrmd_op = NULL;
+ op->rapop = NULL;
+ }
op->first_line_ra_stdout[0] = EOS;
lrmd_debug2(LOG_DEBUG, "lrmd_op_destroy: free the op whose address is
%p"
@@ -487,10 +577,7 @@
op->exec_pid = -1;
op->timeout_tag = -1;
op->repeat_timeout_tag = -1;
- op->ra_stdout_fd = -1;
- op->ra_stderr_fd = -1;
- op->ra_stdout_gsource = NULL;
- op->ra_stderr_gsource = NULL;
+ op->rapop = NULL;
op->first_line_ra_stdout[0] = EOS;
op->t_recv = time_longclock();
++lrm_objectstats.opcount;
@@ -516,16 +603,12 @@
* Be sure and care of this situation when using this function.
*/
/* Do a "deep" copy of the message structure */
+ ret->rapop = NULL;
ret->msg = ha_msg_copy(op->msg);
ret->rsc_id = cl_strdup(op->rsc_id);
- ret->ra_stdout_fd = -1;
- ret->ra_stderr_fd = -1;
- ret->ra_stdout_gsource = NULL;
- ret->ra_stderr_gsource = NULL;
ret->timeout_tag = -1;
ret->repeat_timeout_tag = -1;
ret->exec_pid = -1;
- memset(ret->first_line_ra_stdout, 0, sizeof(ret->first_line_ra_stdout));
ret->is_copy = TRUE;
return ret;
}
@@ -1820,7 +1903,7 @@
CHECK_ALLOCATED(client, "client", HA_FAIL);
CHECK_ALLOCATED(msg, "message", HA_FAIL);
- lrmd_debug(LOG_DEBUG
+ lrmd_debug(LOG_DEBUG
, "on_msg_get_rsc_classes:client [%d] wants to get rsc classes"
, client->pid);
@@ -2993,7 +3076,7 @@
GHashTable* op_params = NULL;
unsigned long t_stay_in_list = 0;
lrmd_rsc_t* rsc = NULL;
- int flag;
+ ra_pipe_op_t * rapop;
CHECK_ALLOCATED(op, "op", HA_FAIL);
rsc = (lrmd_rsc_t*)lookup_rsc(op->rsc_id);
@@ -3049,55 +3132,18 @@
default: /* Parent */
NewTrackedProc(pid, 1, PT_LOGNONE, op,
&ManagedChildTrackOps);
+
close(stdout_fd[1]);
close(stderr_fd[1]);
- /*
- * No any obviouse proof of lrmd hang in pipe read yet.
- * Bug 475 may be a duplicate of bug 499.
- * Anyway, via test, it's proved that NOBLOCK read will
- * obviously reduce the RA execution time (bug 553).
- */
- /* Let the read operation be NONBLOCK */
- if ((flag = fcntl(stdout_fd[0], F_GETFL)) >= 0) {
- if (fcntl(stdout_fd[0], F_SETFL,
flag|O_NONBLOCK) < 0) {
- cl_perror("%s::%d: fcntl", __FUNCTION__
- , __LINE__);
- }
- } else {
- cl_perror("%s::%d: fcntl", __FUNCTION__,
__LINE__);
- }
- if ((flag = fcntl(stderr_fd[0], F_GETFL)) >= 0) {
- if (fcntl(stderr_fd[0], F_SETFL,
flag|O_NONBLOCK) < 0) {
- cl_perror("%s::%d: fcntl", __FUNCTION__
- , __LINE__);
- }
- } else {
- cl_perror("%s::%d: fcntl", __FUNCTION__,
__LINE__);
- }
-
- op->ra_stdout_fd = stdout_fd[0];
- op->ra_stderr_fd = stderr_fd[0];
- if (op->ra_stdout_fd <= STDERR_FILENO) {
- lrmd_log(LOG_ERR
- , "%s: invalid stdout fd [%d]"
- , __FUNCTION__, op->ra_stdout_fd);
- }
- if (op->ra_stderr_fd <= STDERR_FILENO) {
- lrmd_log(LOG_ERR
- , "%s: invalid stderr fd [%d]"
- , __FUNCTION__, op->ra_stderr_fd);
- }
-
- op->ra_stdout_gsource = G_main_add_fd(G_PRIORITY_HIGH
- , stdout_fd[0], FALSE, handle_pipe_ra_stdout
- , op, NULL);
- op->ra_stderr_gsource = G_main_add_fd(G_PRIORITY_HIGH
- , stderr_fd[0], FALSE, handle_pipe_ra_stderr
- , op, NULL);
-
+ rapop = ra_pipe_op_new(stdout_fd[0], stderr_fd[0], op);
+ op->rapop = rapop;
op->exec_pid = pid;
-
+
return_to_dropped_privs();
+
+ if ( rapop == NULL) {
+ return HA_FAIL;
+ }
return HA_OK;
case 0: /* Child */
@@ -3219,10 +3265,25 @@
}
op_type = ha_msg_value(op->msg, F_LRM_OP);
- if (op->first_line_ra_stdout[0] == EOS && op->ra_stdout_fd >= 0) {
- handle_pipe_ra_stdout(-1, op);
+
+ if ( (NULL == strchr(op->first_line_ra_stdout, '\n'))
+ && (0==STRNCMP_CONST(rsc->class, "heartbeat"))
+ && ( (0==STRNCMP_CONST(op_type, "monitor"))
+ ||(0==STRNCMP_CONST(op_type, "status"))) ) {
+ if ( ( op->rapop != NULL )
+ && (op->rapop->ra_stdout_fd >= 0) ) {
+ handle_pipe_ra_stdout(op->rapop->ra_stdout_fd
+ , op->rapop);
+ } else {
+ lrmd_log(LOG_WARNING, "There is something wrong: the "
+ "first line isn't read in. Maybe the heartbeat "
+ "does not ouput string correctly for status "
+ "operation. Or the code (myself) is wrong.");
+ }
}
- rc = RAExec->map_ra_retvalue(exitcode, op_type,
op->first_line_ra_stdout);
+
+ rc = RAExec->map_ra_retvalue(exitcode, op_type
+ , op->first_line_ra_stdout);
if (rc != EXECRA_OK || debug_level > 0) {
if (signo != 0) {
lrmd_log(rc == EXECRA_OK ? LOG_DEBUG : LOG_WARNING
@@ -3354,120 +3415,110 @@
return rsc;
}
-#if 0
static void
destroy_pipe_ra_stdout(gpointer user_data)
{
- lrmd_op_t* op = (lrmd_op_t *)user_data;
+ ra_pipe_op_t * rapop = (ra_pipe_op_t *)user_data;
- if (!cl_is_allocated(op)) {
- lrmd_log(LOG_CRIT, "%s:%d: Unallocated op 0x%lx!!"
- , __FUNCTION__, __LINE__
- , (unsigned long)op);
- return;
- }
- if (op->ra_stdout_fd > STDERR_FILENO) {
- close(op->ra_stdout_fd);
- op->ra_stdout_fd = -1;
- }else if (op->ra_stdout_fd >= 0) {
- lrmd_log(LOG_ERR, "%s:Attempt to close file descriptor %d"
- , __FUNCTION__, op->ra_stdout_fd);
+ CHECK_ALLOCATED(rapop, "ra_pipe_op",);
+ if (rapop->ra_stderr_fd < 0) {
+ ra_pipe_op_destroy(rapop);
}
-
}
static void
destroy_pipe_ra_stderr(gpointer user_data)
{
- lrmd_op_t* op = (lrmd_op_t *)user_data;
+ ra_pipe_op_t * rapop = (ra_pipe_op_t *)user_data;
- if (!cl_is_allocated(op)) {
- lrmd_log(LOG_CRIT, "%s:%d: Unallocated op 0x%lx!!"
- , __FUNCTION__, __LINE__
- , (unsigned long)op);
- return;
- }
- if (op->ra_stderr_fd > STDERR_FILENO) {
- close(op->ra_stderr_fd);
- op->ra_stderr_fd = -1;
- }else if (op->ra_stderr_fd >= 0) {
- lrmd_log(LOG_ERR, "%s:Attempt to close file descriptor %d"
- , __FUNCTION__, op->ra_stderr_fd);
+ CHECK_ALLOCATED(rapop, "ra_pipe_op",);
+ if (rapop->ra_stdout_fd < 0) {
+ ra_pipe_op_destroy(rapop);
}
}
-#endif
static gboolean
handle_pipe_ra_stdout(int fd, gpointer user_data)
{
gboolean rc = TRUE;
- lrmd_op_t* op = (lrmd_op_t *)user_data;
- const char* op_type = NULL;
+ ra_pipe_op_t * rapop = (ra_pipe_op_t *)user_data;
char * data = NULL;
+ lrmd_op_t* lrmd_op = NULL;
- if (op == NULL) {
- lrmd_log(LOG_ERR, "%s:%d: op==NULL.", __FUNCTION__, __LINE__);
- return FALSE;
- }
- if (!cl_is_allocated(op)) {
- lrmd_log(LOG_CRIT, "%s:%d: Unallocated op 0x%lx!!"
- , __FUNCTION__, __LINE__
- , (unsigned long)op);
- return FALSE;
+ CHECK_ALLOCATED(rapop, "ra_pipe_op", FALSE);
- }
-
- if (fd < 0) {
- fd = op->ra_stdout_fd;
+ if (rapop->lrmd_op == NULL) {
+ lrmd_debug(LOG_DEBUG, "%s:%d: Unallocated lrmd_op 0x%lx!!"
+ , __FUNCTION__, __LINE__
+ , (unsigned long)rapop->lrmd_op);
+ } else {
+ lrmd_op = rapop->lrmd_op;
}
if (fd <= STDERR_FILENO) {
- lrmd_log(LOG_CRIT
- , "%s:%d: Attempt to read from closed/invalid file
descriptor %d."
+ lrmd_log(LOG_CRIT, "%s:%d: Attempt to read from "
+ "closed/invalid file descriptor %d."
, __FUNCTION__, __LINE__, fd);
- lrmd_op_dump(op, "op w/closed fd");
return FALSE;
}
- if (0 != read_pipe(fd, &data, op)) {
- if ( NULL != op->ra_stdout_gsource) {
- G_main_del_fd(op->ra_stdout_gsource);
- op->ra_stdout_gsource = NULL;
- }
+ if (0 != read_pipe(fd, &data, rapop)) {
+ /* error or reach the EOF */
if (fd > STDERR_FILENO) {
close(fd);
- if (fd == op->ra_stdout_fd) {
- op->ra_stdout_fd = -1;
+ if (fd == rapop->ra_stdout_fd) {
+ rapop->ra_stdout_fd = -1;
}
}
+ if ( NULL != rapop->ra_stdout_gsource) {
+ /* Don't try to optimize it */
+ GFDSource * tmp;
+ tmp = rapop->ra_stdout_gsource;
+ rapop->ra_stdout_gsource = NULL;
+ G_main_del_fd(tmp);
+ }
rc = FALSE;
}
- if (data!=NULL) {
- op_type = ha_msg_value(op->msg, F_LRM_OP);
- if ( (0==STRNCMP_CONST(op_type, "meta-data"))
- ||(0==STRNCMP_CONST(op_type, "monitor"))
- ||(0==STRNCMP_CONST(op_type, "status")) ) {
+ if ( data!=NULL ) {
+ if ( (0==STRNCMP_CONST(rapop->op_type, "meta-data"))
+ ||(0==STRNCMP_CONST(rapop->op_type, "monitor"))
+ ||(0==STRNCMP_CONST(rapop->op_type, "status")) ) {
lrmd_debug(LOG_DEBUG, "RA output: (%s:%s:stdout) %s"
- , op->rsc_id, op_type, data);
+ , rapop->rsc_id, rapop->op_type, data);
} else {
lrmd_log(LOG_INFO, "RA output: (%s:%s:stdout) %s"
- , op->rsc_id, op_type, data);
+ , rapop->rsc_id, rapop->op_type, data);
}
+
/*
- * This code isn't correct - there is no idea of a "line"
- * in the code as it's presently written...
- * It produces erratic and hard-to read messages in the logs.
- * Under the right circumstances, it will cause errors in
- * interpreting 'heartbeat' style resource agents. FIXME
- * It's not obvious how the meta-data (which is _many_
- * lines long) is handled by this...
- * I suspect multi-line metadata hasn't been tested :-(.
+ * This code isn't good enough, it produces erratic and hard-to
+ * read messages in the logs. But this does not affect the
+ * function correctness, since the first line output is ensured
+ * to be collected into the buffer completely.
+ * Anyway, the meta-data (which is _many_ lines long) can be
+ * handled by another function, see raexec.h
*/
- if (op->first_line_ra_stdout[0] == EOS) {
- strncpy(op->first_line_ra_stdout, data
- , sizeof(op->first_line_ra_stdout) - 1);
+ if ( (rapop->first_line_read == FALSE)
+ && (0==STRNCMP_CONST(rapop->rsc_class, "heartbeat"))
+ && ( lrmd_op != NULL )
+ && ( (0==STRNCMP_CONST(rapop->op_type, "monitor"))
+ ||(0==STRNCMP_CONST(rapop->op_type, "status")) )) {
+ if (lrmd_op != NULL) {
+ strncat(lrmd_op->first_line_ra_stdout, data
+ , sizeof(lrmd_op->first_line_ra_stdout) -
+ strlen(lrmd_op->first_line_ra_stdout)-1);
+ if (strchr(lrmd_op->first_line_ra_stdout, '\n')
+ != NULL) {
+ rapop->first_line_read = TRUE;
+ }
+ } else {
+ lrmd_log(LOG_CRIT
+ , "Before read the first line, the RA "
+ "execution child quitted and waited.");
+ }
}
+
g_free(data);
}
@@ -3478,43 +3529,44 @@
handle_pipe_ra_stderr(int fd, gpointer user_data)
{
gboolean rc = TRUE;
- lrmd_op_t* op = (lrmd_op_t *)user_data;
- const char* op_type = NULL;
char * data = NULL;
+ ra_pipe_op_t * rapop = (ra_pipe_op_t *)user_data;
- if (!cl_is_allocated(op)) {
- lrmd_log(LOG_CRIT, "%s:%d: Unallocated op 0x%lx!!"
- , __FUNCTION__, __LINE__
- , (unsigned long)op);
- return FALSE;
- }
+ CHECK_ALLOCATED(rapop, "ra_pipe_op", FALSE);
if (fd <= STDERR_FILENO) {
- lrmd_log(LOG_CRIT
- , "%s:%d: Attempt to read from closed/invalid file
descriptor %d."
+ lrmd_log(LOG_CRIT, "%s:%d: Attempt to read from "
+ " closed/invalid file descriptor %d."
, __FUNCTION__, __LINE__, fd);
- lrmd_op_dump(op, "op w/closed fd");
return FALSE;
}
- if (0 != read_pipe(fd, &data, op)) {
- if ( NULL != op->ra_stderr_gsource) {
- G_main_del_fd(op->ra_stderr_gsource);
- op->ra_stderr_gsource = NULL;
- }
- if (fd >= STDERR_FILENO) {
+ if (0 != read_pipe(fd, &data, rapop)) {
+ /* error or reach the EOF */
+ if (fd > STDERR_FILENO) {
close(fd);
- if (fd == op->ra_stderr_fd) {
- op->ra_stderr_fd = -1;
+ if (fd == rapop->ra_stderr_fd) {
+ rapop->ra_stderr_fd = -1;
}
}
+ if ( NULL != rapop->ra_stderr_gsource) {
+ /*
+ * Don't try to optimize it.
+ * G_main_del_fd will trigger
+ * destroy_pipe_ra_stderr
+ * ra_pipe_op_destroy
+ */
+ GFDSource * tmp;
+ tmp = rapop->ra_stderr_gsource;
+ rapop->ra_stderr_gsource = NULL;
+ G_main_del_fd(tmp);
+ }
rc = FALSE;
}
if (data!=NULL) {
- op_type = ha_msg_value(op->msg, F_LRM_OP);
- lrmd_log(LOG_INFO, "RA output: (%s:%s:stderr) %s", op->rsc_id
- , op_type, data);
+ lrmd_log(LOG_INFO, "RA output: (%s:%s:stderr) %s"
+ , rapop->rsc_id, rapop->op_type, data);
g_free(data);
}
@@ -3528,25 +3580,32 @@
char buffer[BUFFLEN];
int readlen;
GString * gstr_tmp;
- const char* op_type = NULL;
int rc = 0;
- lrmd_rsc_t* rsc = NULL;
- lrmd_op_t* op = (lrmd_op_t *)user_data;
+ lrmd_op_t * op = NULL;
+ ra_pipe_op_t * rapop = (ra_pipe_op_t *)user_data;
lrmd_debug3(LOG_DEBUG, "%s begin.", __FUNCTION__);
- if (!cl_is_allocated(op)) {
- lrmd_log(LOG_CRIT, "%s:%d: Unallocated op 0x%lx!!"
+ CHECK_ALLOCATED(rapop, "ra_pipe_op", FALSE);
+
+ op = (lrmd_op_t *)rapop->lrmd_op;
+ if (NULL == op) {
+ lrmd_debug2(LOG_DEBUG, "%s:%d: Unallocated lrmd_op 0x%lx!!"
, __FUNCTION__, __LINE__
, (unsigned long)op);
- return FALSE;
}
+
*data = NULL;
gstr_tmp = g_string_new("");
do {
errno = 0;
readlen = read(fd, buffer, BUFFLEN - 1);
+ if (NULL == op) {
+ lrmd_debug2(LOG_NOTICE
+ , "read's ret: %d when lrmd_op finished"
+ , readlen);
+ }
if ( readlen > 0 ) {
buffer[readlen] = EOS;
g_string_append(gstr_tmp, buffer);
@@ -3569,7 +3628,13 @@
cl_perror("%s:%d read error: fd %d errno=%d"
, __FUNCTION__, __LINE__
, fd, errno);
- lrmd_op_dump(op, "op w/bad errno");
+ if (NULL != op) {
+ lrmd_op_dump(op, "op w/bad errno");
+ } else {
+ lrmd_log(LOG_NOTICE
+ , "%s::%d: lrmd_op has been freed"
+ , __FUNCTION__, __LINE__);
+ }
break;
case EBADF:
@@ -3577,22 +3642,33 @@
, "%s:%d"
" Attempt to read from closed file descriptor %d."
, __FUNCTION__, __LINE__, fd);
- lrmd_op_dump(op, "op w/closed fd");
+ if (NULL != op) {
+ lrmd_op_dump(op, "op w/bad errno");
+ } else {
+ lrmd_log(LOG_NOTICE
+ , "%s::%d: lrmd_op has been freed"
+ , __FUNCTION__, __LINE__);
+ }
break;
case EAGAIN:
- rsc = lookup_rsc(op->rsc_id);
- op_type = ha_msg_value(op->msg, F_LRM_OP);
+ if (NULL==op) {
+ lrmd_log(LOG_WARNING
+ , "%s::%d: Shouldn't come here: "
+ "lrmd_op = NULL"
+ , __FUNCTION__, __LINE__);
+ break;
+ }
lrmd_log(LOG_ERR, "RA %s:%s:%s (process %d) failed to "
"redirect %s for its background child (daemon) "
"processes. This will likely cause those "
"processes to die mysteriously at some later "
"time (terminated by signal SIGPIPE)."
- , (rsc!=NULL)?rsc->class:"<NULL>"
- , (rsc!=NULL)?rsc->type:"<NULL>"
- , (op_type!=NULL)?op_type:"<NULL>"
+ , rapop->rsc_class
+ , rapop->rsc_id
+ , rapop->op_type
, op->exec_pid
- , (op->ra_stdout_fd==fd)?"stdout":"stderr");
+ , (rapop->ra_stdout_fd==fd)?"stdout":"stderr");
}
}
@@ -3698,6 +3774,10 @@
}
/*
* $Log: lrmd.c,v $
+ * Revision 1.217 2006/03/13 08:35:13 sunjd
+ * Reworkcl with new implementation for
+ * bug756: All output from resource agents should be logged
+ *
* Revision 1.216 2006/02/10 06:50:23 alan
* Removed destroy_pipe_ra_* functions.
*
@@ -4335,3 +4415,4 @@
* Added missing Id and Log
*
*/
+
------------------------------
_______________________________________________
Linux-ha-cvs mailing list
[email protected]
http://lists.community.tummy.com/mailman/listinfo/linux-ha-cvs
End of Linux-ha-cvs Digest, Vol 28, Issue 24
********************************************
