[Ocfs2-commits] mfasheh commits r2528 -
branches/ocfs2-1.0/fs/ocfs2/cluster
svn-commits at oss.oracle.com
svn-commits at oss.oracle.com
Thu Aug 18 19:24:46 CDT 2005
Author: mfasheh
Signed-off-by: jlbec
Date: 2005-08-18 19:24:43 -0500 (Thu, 18 Aug 2005)
New Revision: 2528
Modified:
branches/ocfs2-1.0/fs/ocfs2/cluster/heartbeat.c
branches/ocfs2-1.0/fs/ocfs2/cluster/heartbeat.h
branches/ocfs2-1.0/fs/ocfs2/cluster/nodemanager.c
branches/ocfs2-1.0/fs/ocfs2/cluster/tcp_internal.h
Log:
* backport r2527 from trunk:
-Make heartbeat dead iterations tunable. This will help sites
who are getting heartbeat timeouts on very busy systems.
Signed-off-by: jlbec
Modified: branches/ocfs2-1.0/fs/ocfs2/cluster/heartbeat.c
===================================================================
--- branches/ocfs2-1.0/fs/ocfs2/cluster/heartbeat.c 2005-08-19 00:21:47 UTC (rev 2527)
+++ branches/ocfs2-1.0/fs/ocfs2/cluster/heartbeat.c 2005-08-19 00:24:43 UTC (rev 2528)
@@ -32,6 +32,7 @@
#include <linux/configfs.h>
#include <linux/random.h>
#include <linux/crc32.h>
+#include <linux/time.h>
#include "heartbeat.h"
#include "tcp.h"
@@ -69,6 +70,23 @@
#define O2HB_DEFAULT_BLOCK_BITS 9
+unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD;
+
+/* Only sets a new threshold if there are no active regions.
+ *
+ * No locking or otherwise interesting code is required for reading
+ * o2hb_dead_threshold as it can't change once regions are active and
+ * it's not interesting to anyone until then anyway. */
+void o2hb_dead_threshold_set(unsigned int threshold)
+{
+ if (threshold > O2HB_MIN_DEAD_THRESHOLD) {
+ spin_lock(&o2hb_live_lock);
+ if (list_empty(&o2hb_all_regions))
+ o2hb_dead_threshold = threshold;
+ spin_unlock(&o2hb_live_lock);
+ }
+}
+
struct o2hb_node_event {
struct list_head hn_item;
enum o2hb_callback_type hn_event_type;
@@ -118,7 +136,6 @@
char hr_dev_name[BDEVNAME_SIZE];
unsigned int hr_timeout_ms;
- unsigned int hr_dead_iter;
/* randomized as the region goes up and down so that a node
* recognizes a node going up and down in one iteration */
@@ -713,7 +730,7 @@
/* live nodes only go dead after enough consequtive missed
* samples.. reset the missed counter whenever we see
* activity */
- if (slot->ds_equal_samples >= reg->hr_dead_iter || gen_changed) {
+ if (slot->ds_equal_samples >= o2hb_dead_threshold || gen_changed) {
mlog(ML_HEARTBEAT, "Node %d left my region\n",
slot->ds_node_num);
@@ -1086,15 +1103,14 @@
static void o2hb_init_region_params(struct o2hb_region *reg)
{
reg->hr_slots_per_page = PAGE_CACHE_SIZE >> reg->hr_block_bits;
- reg->hr_dead_iter = O2HB_DEAD_THRESHOLD;
reg->hr_timeout_ms = O2HB_REGION_TIMEOUT_MS;
mlog(ML_HEARTBEAT, "hr_start_block = %llu, hr_blocks = %u\n",
reg->hr_start_block, reg->hr_blocks);
mlog(ML_HEARTBEAT, "hr_block_bytes = %u, hr_block_bits = %u\n",
reg->hr_block_bytes, reg->hr_block_bits);
- mlog(ML_HEARTBEAT, "hr_timeout_ms = %u, hr_dead_iter = %u\n",
- reg->hr_timeout_ms, reg->hr_dead_iter);
+ mlog(ML_HEARTBEAT, "hr_timeout_ms = %u\n", reg->hr_timeout_ms);
+ mlog(ML_HEARTBEAT, "dead threshold = %u\n", o2hb_dead_threshold);
}
static int o2hb_map_slot_data(struct o2hb_region *reg)
Modified: branches/ocfs2-1.0/fs/ocfs2/cluster/heartbeat.h
===================================================================
--- branches/ocfs2-1.0/fs/ocfs2/cluster/heartbeat.h 2005-08-19 00:21:47 UTC (rev 2527)
+++ branches/ocfs2-1.0/fs/ocfs2/cluster/heartbeat.h 2005-08-19 00:24:43 UTC (rev 2528)
@@ -34,9 +34,14 @@
/* number of changes to be seen as live */
#define O2HB_LIVE_THRESHOLD 2
/* number of equal samples to be seen as dead */
-#define O2HB_DEAD_THRESHOLD 7
+extern unsigned int o2hb_dead_threshold;
+#define O2HB_DEFAULT_DEAD_THRESHOLD 7
+/* Otherwise MAX_WRITE_TIMEOUT will be zero... */
+#define O2HB_MIN_DEAD_THRESHOLD 2
+#define O2HB_MAX_WRITE_TIMEOUT_MS (O2HB_REGION_TIMEOUT_MS * (o2hb_dead_threshold - 1))
-#define O2HB_MAX_WRITE_TIMEOUT_MS (O2HB_REGION_TIMEOUT_MS * (O2HB_DEAD_THRESHOLD - 1))
+/* Always use this to set o2hb_dead_threshold */
+void o2hb_dead_threshold_set(unsigned int threshold);
#define O2HB_CB_MAGIC 0x51d1e4ec
Modified: branches/ocfs2-1.0/fs/ocfs2/cluster/nodemanager.c
===================================================================
--- branches/ocfs2-1.0/fs/ocfs2/cluster/nodemanager.c 2005-08-19 00:21:47 UTC (rev 2527)
+++ branches/ocfs2-1.0/fs/ocfs2/cluster/nodemanager.c 2005-08-19 00:24:43 UTC (rev 2528)
@@ -732,10 +732,12 @@
static struct proc_dir_entry *o2nm_proc;
#define O2NM_VERSION_PROC_NAME "interface_revision"
+#define O2NM_HB_DEAD_THRESHOLD_NAME "hb_dead_threshold"
static void o2nm_remove_proc(struct proc_dir_entry *parent)
{
remove_proc_entry(O2NM_VERSION_PROC_NAME, parent);
+ remove_proc_entry(O2NM_HB_DEAD_THRESHOLD_NAME, parent);
}
static void __exit exit_o2nm(void)
@@ -754,12 +756,12 @@
o2net_exit();
}
-static int o2nm_proc_version(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+static int o2nm_proc_read_uint(char *page, char **start, off_t off,
+ int count, int *eof, unsigned int data)
{
int len;
- len = sprintf(page, "%u\n", O2NM_API_VERSION);
+ len = sprintf(page, "%u\n", data);
if (len < 0)
return len;
@@ -779,6 +781,46 @@
return len;
}
+static int o2nm_proc_version(char *page, char **start, off_t off,
+ int count, int *eof, void *data)
+{
+ return o2nm_proc_read_uint(page, start, off, count, eof,
+ O2NM_API_VERSION);
+}
+
+static int o2nm_proc_threshold(char *page, char **start, off_t off,
+ int count, int *eof, void *data)
+{
+ return o2nm_proc_read_uint(page, start, off, count, eof,
+ o2hb_dead_threshold);
+}
+
+static int o2nm_proc_write_threshold(struct file *file,
+ const char __user *buffer,
+ unsigned long count, void *data)
+{
+ char buf[32];
+ char *p = buf;
+ unsigned long tmp;
+
+ if (count > ARRAY_SIZE(buf) - 1)
+ count = ARRAY_SIZE(buf) - 1;
+
+ if (copy_from_user(buf, buffer, count))
+ return -EFAULT;
+
+ buf[ARRAY_SIZE(buf) - 1] = '\0';
+
+ tmp = simple_strtoul(p, &p, 10);
+ if (!p || (*p && (*p != '\n')))
+ return -EINVAL;
+
+ /* this will validate ranges for us. */
+ o2hb_dead_threshold_set((unsigned int) tmp);
+
+ return count;
+}
+
static int o2nm_init_proc(struct proc_dir_entry *parent)
{
struct proc_dir_entry *p;
@@ -790,6 +832,17 @@
NULL);
if (!p)
return -ENOMEM;
+
+ p = create_proc_read_entry(O2NM_HB_DEAD_THRESHOLD_NAME,
+ S_IFREG | S_IRUGO | S_IWUSR, parent,
+ o2nm_proc_threshold,
+ NULL);
+ if (!p) {
+ remove_proc_entry(O2NM_VERSION_PROC_NAME, parent);
+ return -ENOMEM;
+ }
+ p->write_proc = o2nm_proc_write_threshold;
+
return 0;
}
Modified: branches/ocfs2-1.0/fs/ocfs2/cluster/tcp_internal.h
===================================================================
--- branches/ocfs2-1.0/fs/ocfs2/cluster/tcp_internal.h 2005-08-19 00:21:47 UTC (rev 2527)
+++ branches/ocfs2-1.0/fs/ocfs2/cluster/tcp_internal.h 2005-08-19 00:24:43 UTC (rev 2528)
@@ -35,7 +35,7 @@
/* we're delaying our quorum decision so that heartbeat will have timed
* out truly dead nodes by the time we come around to making decisions
* on their number */
-#define O2NET_QUORUM_DELAY_MS ((O2HB_DEAD_THRESHOLD + 2) * O2HB_REGION_TIMEOUT_MS)
+#define O2NET_QUORUM_DELAY_MS ((o2hb_dead_threshold + 2) * O2HB_REGION_TIMEOUT_MS)
#define O2NET_KEEPALIVE_DELAY_SECS 5
#define O2NET_IDLE_TIMEOUT_SECS 10
More information about the Ocfs2-commits
mailing list