Allow scaling of arc in proportion to pagecache

When multiple filesystems are in use, memory pressure causes arc_cache
to collapse to a minimum. Allow arc_cache to maintain proportional size
even when hit rates are disproportionate. We do this only via evictable
size from the kernel shrinker, thus it's only in effect under memory
pressure.

AKAMAI: zfs: CR 3695072
Reviewed-by: Tim Chase <tim@chase2k.com>
Reviewed-by: Richard Yao <ryao@gentoo.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Debabrata Banerjee <dbanerje@akamai.com>
Closes #6035
This commit is contained in:
Debabrata Banerjee 2017-03-15 21:34:56 -04:00 committed by Brian Behlendorf
parent 4149bf498a
commit 03b60eee78
2 changed files with 37 additions and 2 deletions

View File

@ -699,6 +699,24 @@ log2(fraction of arc to reclaim)
Default value: \fB5\fR.
.RE
.sp
.ne 2
.na
\fBzfs_arc_pc_percent\fR (uint)
.ad
.RS 12n
Percent of pagecache to reclaim arc to
This tunable allows ZFS arc to play more nicely with the kernel's LRU
pagecache. It can guarantee that the arc size won't collapse under scanning
pressure on the pagecache, yet still allows arc to be reclaimed down to
zfs_arc_min if necessary. This value is specified as percent of pagecache
size (as measured by NR_FILE_PAGES) where that percent may exceed 100. This
only operates during memory pressure/reclaim.
.sp
Default value: \fB0\fR (disabled).
.RE
.sp
.ne 2
.na

View File

@ -319,6 +319,11 @@ static int arc_p_min_shift = 4;
/* log2(fraction of arc to reclaim) */
static int arc_shrink_shift = 7;
/* percent of pagecache to reclaim arc to */
#ifdef _KERNEL
static uint_t zfs_arc_pc_percent = 0;
#endif
/*
* log2(fraction of ARC which must be free to allow growing).
* I.e. If there is less than arc_c >> arc_no_grow_shift free memory,
@ -4377,10 +4382,18 @@ arc_evictable_memory(void)
refcount_count(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]);
uint64_t arc_dirty = MAX((int64_t)arc_size - (int64_t)arc_clean, 0);
if (arc_dirty >= arc_c_min)
/*
* Scale reported evictable memory in proportion to page cache, cap
* at specified min/max.
*/
uint64_t min = (ptob(global_page_state(NR_FILE_PAGES)) / 100) *
zfs_arc_pc_percent;
min = MAX(arc_c_min, MIN(arc_c_max, min));
if (arc_dirty >= min)
return (arc_clean);
return (MAX((int64_t)arc_size - (int64_t)arc_c_min, 0));
return (MAX((int64_t)arc_size - (int64_t)min, 0));
}
/*
@ -7754,6 +7767,10 @@ MODULE_PARM_DESC(zfs_arc_p_dampener_disable, "disable arc_p adapt dampener");
module_param(zfs_arc_shrink_shift, int, 0644);
MODULE_PARM_DESC(zfs_arc_shrink_shift, "log2(fraction of arc to reclaim)");
module_param(zfs_arc_pc_percent, uint, 0644);
MODULE_PARM_DESC(zfs_arc_pc_percent,
"Percent of pagecache to reclaim arc to");
module_param(zfs_arc_p_min_shift, int, 0644);
MODULE_PARM_DESC(zfs_arc_p_min_shift, "arc_c shift to calc min/max arc_p");