leandrof/linux
1
0
Fork 0
forked from Minki/linux
Code Issues Pull Requests Packages Projects Releases Wiki Activity

staging/lustre/ldlm: high load because of negative timeout

Browse Source 
When the time of LRU resizing exceeds waiting period of
recalculation, the ldlm daemon will keep on resizing without any
interval of rest. That will cause high CPU load.

This patch fixes the problem by setting the recalculation timestamp
after LRU resizing finishes rather than before it. What is more, an
interval of one second is enforced between each recalculation.

Signed-off-by: Li Xi <lixi@ddn.com>
Reviewed-on: http://review.whamcloud.com/11227
Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5415
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Bobi Jam <bobijam@gmail.com>
Reviewed-by: Lai Siyao <lai.siyao@intel.com>
Signed-off-by: Oleg Drokin <oleg.drokin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This commit is contained in:
Li Xi 2015-02-01 21:52:02 -05:00 committed by Greg Kroah-Hartman
parent c00266e369
commit 4d2c7b309d
1 changed files with 26 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
drivers/staging/lustre/lustre/ldlm/ldlm_pool.c
View File

@ -470,6 +470,7 @@ static void ldlm_cli_pool_pop_slv(struct ldlm_pool *pl)
static int ldlm_cli_pool_recalc(struct ldlm_pool *pl)
{
time_t recalc_interval_sec;
int ret;
recalc_interval_sec = get_seconds() - pl->pl_recalc_time;
if (recalc_interval_sec < pl->pl_recalc_period)
@ -490,16 +491,15 @@ static int ldlm_cli_pool_recalc(struct ldlm_pool *pl)
*/
ldlm_cli_pool_pop_slv(pl);
pl->pl_recalc_time = get_seconds();
lprocfs_counter_add(pl->pl_stats, LDLM_POOL_TIMING_STAT,
recalc_interval_sec);
spin_unlock(&pl->pl_lock);
/*
* Do not cancel locks in case lru resize is disabled for this ns.
*/
if (!ns_connect_lru_resize(ldlm_pl2ns(pl)))
return 0;
if (!ns_connect_lru_resize(ldlm_pl2ns(pl))) {
ret = 0;
goto out;
}
/*
* In the time of canceling locks on client we do not need to maintain
@ -507,7 +507,19 @@ static int ldlm_cli_pool_recalc(struct ldlm_pool *pl)
* It may be called when SLV has changed much, this is why we do not
* take into account pl->pl_recalc_time here.
*/
return ldlm_cancel_lru(ldlm_pl2ns(pl), 0, LCF_ASYNC, LDLM_CANCEL_LRUR);
ret = ldlm_cancel_lru(ldlm_pl2ns(pl), 0, LCF_ASYNC, LDLM_CANCEL_LRUR);
out:
spin_lock(&pl->pl_lock);
/*
* Time of LRU resizing might be longer than period,
* so update after LRU resizing rather than before it.
*/
pl->pl_recalc_time = get_seconds();
lprocfs_counter_add(pl->pl_stats, LDLM_POOL_TIMING_STAT,
recalc_interval_sec);
spin_unlock(&pl->pl_lock);
return ret;
}
/**
@ -591,6 +603,14 @@ int ldlm_pool_recalc(struct ldlm_pool *pl)
}
recalc_interval_sec = pl->pl_recalc_time - get_seconds() +
pl->pl_recalc_period;
if (recalc_interval_sec <= 0) {
/* Prevent too frequent recalculation. */
CDEBUG(D_DLMTRACE, "Negative interval(%ld), "
"too short period(%ld)",
recalc_interval_sec,
pl->pl_recalc_period);
recalc_interval_sec = 1;
}
return recalc_interval_sec;
}