c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2026-05-26 12:12:13 +03:00
Code Issues Packages Projects Releases Wiki Activity

Illumos #1051: zfs should handle imbalanced luns

Browse Source 
Today zfs tries to allocate blocks evenly across all devices.
This means when devices are imbalanced zfs will use lots of
CPU searching for space on devices which tend to be pretty
full.  It should instead fail quickly on the full LUNs and
move onto devices which have more availability.

Reviewed by: Eric Schrock <Eric.Schrock@delphix.com>
Reviewed by: Matt Ahrens <Matt.Ahrens@delphix.com>
Reviewed by: Adam Leventhal <Adam.Leventhal@delphix.com>
Reviewed by: Albert Lee <trisk@nexenta.com>
Reviewed by: Gordon Ross <gwr@nexenta.com>
Approved by: Garrett D'Amore <garrett@nexenta.com>

References to Illumos issue and patch:
- https://www.illumos.org/issues/510
- https://github.com/illumos/illumos-gate/commit/5ead3ed965

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #340
This commit is contained in:
George Wilson
2011-07-26 12:08:52 -07:00
committed by Brian Behlendorf
parent bb939d1085
commit 6d974228ef
8 changed files with 123 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files
module/zfs/zio.c
+21 -1
View File
@@ -20,6 +20,7 @@
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
@@ -79,6 +80,7 @@ int zio_delay_max = ZIO_DELAY_MAX;
#ifdef _KERNEL
extern vmem_t *zio_alloc_arena;
#endif
extern int zfs_mg_alloc_failures;
/*
* An allocating zio is one that either currently has the DVA allocate
@@ -158,6 +160,12 @@ 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.
*/
zfs_mg_alloc_failures = MAX((3 * max_ncpus / 2), 8);
zio_inject_init();
}
@@ -2151,6 +2159,7 @@ zio_dva_allocate(zio_t *zio)
metaslab_class_t *mc = spa_normal_class(spa);
blkptr_t *bp = zio->io_bp;
int error;
int flags = 0;
if (zio->io_gang_leader == NULL) {
ASSERT(zio->io_child_type > ZIO_CHILD_GANG);
@@ -2163,10 +2172,21 @@ zio_dva_allocate(zio_t *zio)
ASSERT3U(zio->io_prop.zp_copies, <=, spa_max_replication(spa));
ASSERT3U(zio->io_size, ==, BP_GET_PSIZE(bp));
/*
* The dump device does not support gang blocks so allocation on
* behalf of the dump device (i.e. ZIO_FLAG_NODATA) must avoid
* the "fast" gang feature.
*/
flags |= (zio->io_flags & ZIO_FLAG_NODATA) ? METASLAB_GANG_AVOID : 0;
flags |= (zio->io_flags & ZIO_FLAG_GANG_CHILD) ?
METASLAB_GANG_CHILD : 0;
error = metaslab_alloc(spa, mc, zio->io_size, bp,
zio->io_prop.zp_copies, zio->io_txg, NULL, 0);
zio->io_prop.zp_copies, zio->io_txg, NULL, flags);
if (error) {
spa_dbgmsg(spa, "%s: metaslab allocation failure: zio %p, "
"size %llu, error %d", spa_name(spa), zio, zio->io_size,
error);
if (error == ENOSPC && zio->io_size > SPA_MINBLOCKSIZE)
return (zio_write_gang_block(zio));
zio->io_error = error;