d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

MFV r264668:

Browse Source 
4754 io issued to near-full luns even after setting noalloc threshold
4755 mg_alloc_failures is no longer needed

illumos/illumos@b6240e830b

MFC after:	2 weeks
This commit is contained in:
delphij 2014-04-18 22:04:58 +00:00
parent aff2f85dda
commit b281166740
3 changed files with 7 additions and 57 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

51
sys/cddl/contrib/opensolaris/uts/common/fs/zfs/metaslab.c
View File

@ -41,7 +41,7 @@ SYSCTL_NODE(_vfs_zfs, OID_AUTO, metaslab, CTLFLAG_RW, 0, "ZFS metaslab");
* avoid having to load lots of space_maps in a given txg. There are,
* however, some cases where we want to avoid "fast" ganging and instead
* we want to do an exhaustive search of all metaslabs on this device.
* Currently we don't allow any gang, zil, or dump device related allocations
* Currently we don't allow any gang, slog, or dump device related allocations
* to "fast" gang.
*/
#define CAN_FASTGANG(flags) \
@ -73,18 +73,6 @@ SYSCTL_INT(_vfs_zfs, OID_AUTO, condense_pct, CTLFLAG_RWTUN,
"Condense on-disk spacemap when it is more than this many percents"
" of in-memory counterpart");
/*
* This value defines the number of allowed allocation failures per vdev.
* If a device reaches this threshold in a given txg then we consider skipping
* allocations on that device. The value of zfs_mg_alloc_failures is computed
* in zio_init() unless it has been overridden in /etc/system.
*/
int zfs_mg_alloc_failures = 0;
TUNABLE_INT("vfs.zfs.mg_alloc_failures", &zfs_mg_alloc_failures);
SYSCTL_INT(_vfs_zfs, OID_AUTO, mg_alloc_failures, CTLFLAG_RWTUN,
&zfs_mg_alloc_failures, 0,
"Number of allowed allocation failures per vdev");
/*
* The zfs_mg_noalloc_threshold defines which metaslab groups should
* be eligible for allocation. The value is defined as a percentage of
@ -1707,10 +1695,7 @@ metaslab_sync_done(metaslab_t *msp, uint64_t txg)
void
metaslab_sync_reassess(metaslab_group_t *mg)
{
int64_t failures = mg->mg_alloc_failures;
metaslab_group_alloc_update(mg);
atomic_add_64(&mg->mg_alloc_failures, -failures);
/*
* Preload the next potential metaslabs
@ -1737,7 +1722,7 @@ metaslab_distance(metaslab_t *msp, dva_t *dva)
static uint64_t
metaslab_group_alloc(metaslab_group_t *mg, uint64_t psize, uint64_t asize,
uint64_t txg, uint64_t min_distance, dva_t *dva, int d, int flags)
uint64_t txg, uint64_t min_distance, dva_t *dva, int d)
{
spa_t *spa = mg->mg_vd->vdev_spa;
metaslab_t *msp = NULL;
@ -1764,10 +1749,9 @@ metaslab_group_alloc(metaslab_group_t *mg, uint64_t psize, uint64_t asize,
spa_dbgmsg(spa, "%s: failed to meet weight "
"requirement: vdev %llu, txg %llu, mg %p, "
"msp %p, psize %llu, asize %llu, "
"failures %llu, weight %llu",
spa_name(spa), mg->mg_vd->vdev_id, txg,
mg, msp, psize, asize,
mg->mg_alloc_failures, msp->ms_weight);
"weight %llu", spa_name(spa),
mg->mg_vd->vdev_id, txg,
mg, msp, psize, asize, msp->ms_weight);
mutex_exit(&mg->mg_lock);
return (-1ULL);
}
@ -1799,27 +1783,6 @@ metaslab_group_alloc(metaslab_group_t *mg, uint64_t psize, uint64_t asize,
mutex_enter(&msp->ms_lock);
/*
* If we've already reached the allowable number of failed
* allocation attempts on this metaslab group then we
* consider skipping it. We skip it only if we're allowed
* to "fast" gang, the physical size is larger than
* a gang block, and we're attempting to allocate from
* the primary metaslab.
*/
if (mg->mg_alloc_failures > zfs_mg_alloc_failures &&
CAN_FASTGANG(flags) && psize > SPA_GANGBLOCKSIZE &&
activation_weight == METASLAB_WEIGHT_PRIMARY) {
spa_dbgmsg(spa, "%s: skipping metaslab group: "
"vdev %llu, txg %llu, mg %p, msp[%llu] %p, "
"psize %llu, asize %llu, failures %llu",
spa_name(spa), mg->mg_vd->vdev_id, txg, mg,
msp->ms_id, msp, psize, asize,
mg->mg_alloc_failures);
mutex_exit(&msp->ms_lock);
return (-1ULL);
}
/*
* Ensure that the metaslab we have selected is still
* capable of handling our request. It's possible that
@ -1859,8 +1822,6 @@ metaslab_group_alloc(metaslab_group_t *mg, uint64_t psize, uint64_t asize,
if ((offset = metaslab_block_alloc(msp, asize)) != -1ULL)
break;
atomic_inc_64(&mg->mg_alloc_failures);
metaslab_passivate(msp, metaslab_block_maxsize(msp));
mutex_exit(&msp->ms_lock);
}
@ -2015,7 +1976,7 @@ top:
ASSERT(P2PHASE(asize, 1ULL << vd->vdev_ashift) == 0);
offset = metaslab_group_alloc(mg, psize, asize, txg, distance,
dva, d, flags);
dva, d);
if (offset != -1ULL) {
/*
* If we've just selected this metaslab group,

3
sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/metaslab_impl.h
View File

@ -24,7 +24,7 @@
*/
/*
* Copyright (c) 2013 by Delphix. All rights reserved.
* Copyright (c) 2011, 2014 by Delphix. All rights reserved.
*/
#ifndef _SYS_METASLAB_IMPL_H
@ -58,7 +58,6 @@ struct metaslab_group {
kmutex_t mg_lock;
avl_tree_t mg_metaslab_tree;
uint64_t mg_aliquot;
uint64_t mg_alloc_failures;
boolean_t mg_allocatable; /* can we allocate? */
uint64_t mg_free_capacity; /* percentage free */
int64_t mg_bias;

10
sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zio.c
View File

@ -90,7 +90,6 @@ kmem_cache_t *zio_data_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
#ifdef _KERNEL
extern vmem_t *zio_alloc_arena;
#endif
extern int zfs_mg_alloc_failures;
/*
* The following actions directly effect the spa's sync-to-convergence logic.
@ -206,15 +205,6 @@ zio_init(void)
}
out:
/*
* The zio write taskqs have 1 thread per cpu, allow 1/2 of the taskqs
* to fail 3 times per txg or 8 failures, whichever is greater.
*/
if (zfs_mg_alloc_failures == 0)
zfs_mg_alloc_failures = MAX((3 * max_ncpus / 2), 8);
else if (zfs_mg_alloc_failures < 8)
zfs_mg_alloc_failures = 8;
zio_inject_init();
zio_trim_ksp = kstat_create("zfs", 0, "zio_trim", "misc",