svn commit: r264668 - in vendor-sys/illumos/dist/uts/common/fs/zfs: . sys
Xin LI
delphij at FreeBSD.org
Fri Apr 18 19:43:03 UTC 2014
Author: delphij
Date: Fri Apr 18 19:43:02 2014
New Revision: 264668
URL: http://svnweb.freebsd.org/changeset/base/264668
Log:
4754 io issued to near-full luns even after setting noalloc threshold
4755 mg_alloc_failures is no longer needed
illumos/illumos at b6240e830b871f59c22a3918aebb3b36c872edba
Modified:
vendor-sys/illumos/dist/uts/common/fs/zfs/metaslab.c
vendor-sys/illumos/dist/uts/common/fs/zfs/sys/metaslab_impl.h
vendor-sys/illumos/dist/uts/common/fs/zfs/zio.c
Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/metaslab.c
==============================================================================
--- vendor-sys/illumos/dist/uts/common/fs/zfs/metaslab.c Fri Apr 18 19:41:01 2014 (r264667)
+++ vendor-sys/illumos/dist/uts/common/fs/zfs/metaslab.c Fri Apr 18 19:43:02 2014 (r264668)
@@ -38,7 +38,7 @@
* 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) \
@@ -62,14 +62,6 @@ uint64_t metaslab_gang_bang = SPA_MAXBLO
int zfs_condense_pct = 200;
/*
- * 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;
-
-/*
* The zfs_mg_noalloc_threshold defines which metaslab groups should
* be eligible for allocation. The value is defined as a percentage of
* a free space. Metaslab groups that have more free space than
@@ -1611,10 +1603,7 @@ metaslab_sync_done(metaslab_t *msp, uint
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
@@ -1641,7 +1630,7 @@ metaslab_distance(metaslab_t *msp, dva_t
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;
@@ -1668,10 +1657,9 @@ metaslab_group_alloc(metaslab_group_t *m
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);
}
@@ -1704,27 +1692,6 @@ metaslab_group_alloc(metaslab_group_t *m
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
* another thread may have changed the weight while we
@@ -1763,8 +1730,6 @@ metaslab_group_alloc(metaslab_group_t *m
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);
}
@@ -1919,7 +1884,7 @@ 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,
Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/sys/metaslab_impl.h
==============================================================================
--- vendor-sys/illumos/dist/uts/common/fs/zfs/sys/metaslab_impl.h Fri Apr 18 19:41:01 2014 (r264667)
+++ vendor-sys/illumos/dist/uts/common/fs/zfs/sys/metaslab_impl.h Fri Apr 18 19:43:02 2014 (r264668)
@@ -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
@@ -57,7 +57,6 @@ 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;
Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/zio.c
==============================================================================
--- vendor-sys/illumos/dist/uts/common/fs/zfs/zio.c Fri Apr 18 19:41:01 2014 (r264667)
+++ vendor-sys/illumos/dist/uts/common/fs/zfs/zio.c Fri Apr 18 19:43:02 2014 (r264668)
@@ -61,7 +61,6 @@ kmem_cache_t *zio_data_buf_cache[SPA_MAX
#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.
@@ -168,13 +167,6 @@ zio_init(void)
zio_data_buf_cache[c - 1] = zio_data_buf_cache[c];
}
- /*
- * 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);
-
zio_inject_init();
}
More information about the svn-src-all
mailing list