drbd: Implemented the disk-timeout option

When the disk-timeout is active, and it expires for a single request,
we consider the local disk as D_FAILED. Note: With this change,
I made both timeout based state transitions HARD state transitions.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
This commit is contained in:
Philipp Reisner 2011-06-29 14:06:08 +02:00
parent 02ee8f95fa
commit dfa8bedbfe
5 changed files with 31 additions and 14 deletions

View File

@ -1404,6 +1404,9 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os,
/* Here we have the actions that are performed after a
state change. This function might sleep */
if (os.disk <= D_NEGOTIATING && ns.disk > D_NEGOTIATING)
mod_timer(&mdev->request_timer, jiffies + HZ);
nsm.i = -1;
if (ns.susp_nod) {
if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED)
@ -3318,6 +3321,8 @@ static void drbd_delete_device(unsigned int minor)
if (!mdev)
return;
del_timer_sync(&mdev->request_timer);
/* paranoia asserts */
if (mdev->open_cnt != 0)
dev_err(DEV, "open_cnt = %d in %s:%u", mdev->open_cnt,

View File

@ -3803,8 +3803,6 @@ static void drbd_disconnect(struct drbd_conf *mdev)
atomic_set(&mdev->rs_pending_cnt, 0);
wake_up(&mdev->misc_wait);
del_timer(&mdev->request_timer);
/* make sure syncer is stopped and w_resume_next_sg queued */
del_timer_sync(&mdev->resync_timer);
resync_timer_fn((unsigned long)mdev);

View File

@ -1208,13 +1208,19 @@ void request_timer_fn(unsigned long data)
struct drbd_conf *mdev = (struct drbd_conf *) data;
struct drbd_request *req; /* oldest request */
struct list_head *le;
unsigned long et = 0; /* effective timeout = ko_count * timeout */
unsigned long ent = 0, dt = 0, et, nt; /* effective timeout = ko_count * timeout */
if (get_net_conf(mdev)) {
et = mdev->net_conf->timeout*HZ/10 * mdev->net_conf->ko_count;
ent = mdev->net_conf->timeout*HZ/10 * mdev->net_conf->ko_count;
put_net_conf(mdev);
}
if (!et || mdev->state.conn < C_WF_REPORT_PARAMS)
if (get_ldev(mdev)) {
dt = mdev->ldev->dc.disk_timeout * HZ / 10;
put_ldev(mdev);
}
et = min_not_zero(dt, ent);
if (!et || (mdev->state.conn < C_WF_REPORT_PARAMS && mdev->state.disk <= D_FAILED))
return; /* Recurring timer stopped */
spin_lock_irq(&mdev->req_lock);
@ -1227,17 +1233,19 @@ void request_timer_fn(unsigned long data)
le = le->prev;
req = list_entry(le, struct drbd_request, tl_requests);
if (time_is_before_eq_jiffies(req->start_time + et)) {
if (req->rq_state & RQ_NET_PENDING) {
if (ent && req->rq_state & RQ_NET_PENDING) {
if (time_is_before_eq_jiffies(req->start_time + ent)) {
dev_warn(DEV, "Remote failed to finish a request within ko-count * timeout\n");
_drbd_set_state(_NS(mdev, conn, C_TIMEOUT), CS_VERBOSE, NULL);
} else {
dev_warn(DEV, "Local backing block device frozen?\n");
mod_timer(&mdev->request_timer, jiffies + et);
_drbd_set_state(_NS(mdev, conn, C_TIMEOUT), CS_VERBOSE | CS_HARD, NULL);
}
} else {
mod_timer(&mdev->request_timer, req->start_time + et);
}
if (dt && req->rq_state & RQ_LOCAL_PENDING) {
if (time_is_before_eq_jiffies(req->start_time + dt)) {
dev_warn(DEV, "Local backing device failed to meet the disk-timeout\n");
__drbd_chk_io_error(mdev, 1);
}
}
nt = (time_is_before_eq_jiffies(req->start_time + et) ? jiffies : req->start_time) + et;
spin_unlock_irq(&mdev->req_lock);
mod_timer(&mdev->request_timer, nt);
}

View File

@ -48,6 +48,11 @@
#define DRBD_TIMEOUT_MAX 600
#define DRBD_TIMEOUT_DEF 60 /* 6 seconds */
/* If backing disk takes longer than disk_timeout, mark the disk as failed */
#define DRBD_DISK_TIMEOUT_MIN 0 /* 0 = disabled */
#define DRBD_DISK_TIMEOUT_MAX 6000 /* 10 Minutes */
#define DRBD_DISK_TIMEOUT_DEF 0 /* disabled */
/* active connection retries when C_WF_CONNECTION */
#define DRBD_CONNECT_INT_MIN 1
#define DRBD_CONNECT_INT_MAX 120

View File

@ -31,6 +31,7 @@ NL_PACKET(disk_conf, 3,
NL_INTEGER( 56, T_MAY_IGNORE, max_bio_bvecs)
NL_BIT( 57, T_MAY_IGNORE, no_disk_barrier)
NL_BIT( 58, T_MAY_IGNORE, no_disk_drain)
NL_INTEGER( 89, T_MAY_IGNORE, disk_timeout)
)
NL_PACKET(detach, 4,