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

MFV 312436

Browse Source 
6569 large file delete can starve out write ops

  illumos/illumos-gate@ff5177ee8b
  ff5177ee8b

  https://www.illumos.org/issues/6569
    The core issue I've found is that there is no throttle for how many
    deletes get assigned to one TXG. As a results when deleting large files
    we end up filling consecutive TXGs with deletes/frees, then write
    throttling other (more important) ops.

    There is an easy test case for this problem. Try deleting several
    large files (at least 1/2 TB) while you do write ops on the same
    pool. What we've seen is performance of these write ops (let's
    call it sideload I/O) would drop to zero.

    More specifically the problem is that dmu_free_long_range_impl()
    can/will fill up all of the dirty data in the pool "instantly",
    before many of the sideload ops can get in. So sideload
    performance will be impacted until all the files are freed.

    The solution we have tested at Nexenta (with positive results)
    creates a relatively simple throttle for how many "free" ops we let
    into one TXG.

    However this solution exposes other problems that should also be
    addressed. If we are to slow down freeing of data that means one
    has to wait even longer (assuming vnode ref count of 1) to get shell
    back after an rm or for NFS thread to finish the free-ing op.
    To avoid this the proposed solution is to call zfs_inactive() async
    for "large" files. Async freeing then begs for the reclaimed space
    to be accounted for in the zpool's "freeing" prop.

    The other issue with having a longer delete is the inability to
    export/unmount for a longer period of time. The proposed solution
    is to interrupt freeing of blocks when a fs is unmounted.

  Author: Alek Pinchuk <alek@nexenta.com>
  Reviewed by: Matt Ahrens <mahrens@delphix.com>
  Reviewed by: Sanjay Nadkarni <sanjay.nadkarni@nexenta.com>
  Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com>
  Approved by: Dan McDonald <danmcd@omniti.com>

Reviewed by:	avg
Differential Revision:	D9008
This commit is contained in:
Josh Paetzel 2017-01-20 15:01:04 +00:00
commit f2be81e92c
3 changed files with 66 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

59
sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu.c
View File

@ -60,6 +60,16 @@ SYSCTL_DECL(_vfs_zfs);
SYSCTL_INT(_vfs_zfs, OID_AUTO, nopwrite_enabled, CTLFLAG_RDTUN,
&zfs_nopwrite_enabled, 0, "Enable nopwrite feature");
/*
* Tunable to control percentage of dirtied blocks from frees in one TXG.
* After this threshold is crossed, additional dirty blocks from frees
* wait until the next TXG.
* A value of zero will disable this throttle.
*/
uint32_t zfs_per_txg_dirty_frees_percent = 30;
SYSCTL_INT(_vfs_zfs, OID_AUTO, per_txg_dirty_frees_percent, CTLFLAG_RWTUN,
&zfs_per_txg_dirty_frees_percent, 0, "Percentage of dirtied blocks from frees in one txg");
const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = {
{ DMU_BSWAP_UINT8, TRUE, "unallocated" },
{ DMU_BSWAP_ZAP, TRUE, "object directory" },
@ -718,15 +728,25 @@ dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset,
{
uint64_t object_size = (dn->dn_maxblkid + 1) * dn->dn_datablksz;
int err;
uint64_t dirty_frees_threshold;
dsl_pool_t *dp = dmu_objset_pool(os);
if (offset >= object_size)
return (0);
if (zfs_per_txg_dirty_frees_percent <= 100)
dirty_frees_threshold =
zfs_per_txg_dirty_frees_percent * zfs_dirty_data_max / 100;
else
dirty_frees_threshold = zfs_dirty_data_max / 4;
if (length == DMU_OBJECT_END || offset + length > object_size)
length = object_size - offset;
while (length != 0) {
uint64_t chunk_end, chunk_begin;
uint64_t chunk_end, chunk_begin, chunk_len;
uint64_t long_free_dirty_all_txgs = 0;
dmu_tx_t *tx;
chunk_end = chunk_begin = offset + length;
@ -737,9 +757,28 @@ dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset,
ASSERT3U(chunk_begin, >=, offset);
ASSERT3U(chunk_begin, <=, chunk_end);
dmu_tx_t *tx = dmu_tx_create(os);
dmu_tx_hold_free(tx, dn->dn_object,
chunk_begin, chunk_end - chunk_begin);
chunk_len = chunk_end - chunk_begin;
mutex_enter(&dp->dp_lock);
for (int t = 0; t < TXG_SIZE; t++) {
long_free_dirty_all_txgs +=
dp->dp_long_free_dirty_pertxg[t];
}
mutex_exit(&dp->dp_lock);
/*
* To avoid filling up a TXG with just frees wait for
* the next TXG to open before freeing more chunks if
* we have reached the threshold of frees
*/
if (dirty_frees_threshold != 0 &&
long_free_dirty_all_txgs >= dirty_frees_threshold) {
txg_wait_open(dp, 0);
continue;
}
tx = dmu_tx_create(os);
dmu_tx_hold_free(tx, dn->dn_object, chunk_begin, chunk_len);
/*
* Mark this transaction as typically resulting in a net
@ -751,10 +790,18 @@ dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset,
dmu_tx_abort(tx);
return (err);
}
dnode_free_range(dn, chunk_begin, chunk_end - chunk_begin, tx);
mutex_enter(&dp->dp_lock);
dp->dp_long_free_dirty_pertxg[dmu_tx_get_txg(tx) & TXG_MASK] +=
chunk_len;
mutex_exit(&dp->dp_lock);
DTRACE_PROBE3(free__long__range,
uint64_t, long_free_dirty_all_txgs, uint64_t, chunk_len,
uint64_t, dmu_tx_get_txg(tx));
dnode_free_range(dn, chunk_begin, chunk_len, tx);
dmu_tx_commit(tx);
length -= chunk_end - chunk_begin;
length -= chunk_len;
}
return (0);
}

11
sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_pool.c
View File

@ -24,6 +24,7 @@
* Copyright (c) 2013 Steven Hartland. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
* Copyright (c) 2014 Integros [integros.com]
* Copyright 2016 Nexenta Systems, Inc. All rights reserved.
*/
#include <sys/dsl_pool.h>
@ -592,6 +593,16 @@ dsl_pool_sync(dsl_pool_t *dp, uint64_t txg)
*/
dsl_pool_undirty_space(dp, dp->dp_dirty_pertxg[txg & TXG_MASK], txg);
/*
* Update the long range free counter after
* we're done syncing user data
*/
mutex_enter(&dp->dp_lock);
ASSERT(spa_sync_pass(dp->dp_spa) == 1 ||
dp->dp_long_free_dirty_pertxg[txg & TXG_MASK] == 0);
dp->dp_long_free_dirty_pertxg[txg & TXG_MASK] = 0;
mutex_exit(&dp->dp_lock);
/*
* After the data blocks have been written (ensured by the zio_wait()
* above), update the user/group space accounting.

2
sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/dsl_pool.h
View File

@ -21,6 +21,7 @@
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013 by Delphix. All rights reserved.
* Copyright 2016 Nexenta Systems, Inc. All rights reserved.
*/
#ifndef _SYS_DSL_POOL_H
@ -103,6 +104,7 @@ typedef struct dsl_pool {
kcondvar_t dp_spaceavail_cv;
uint64_t dp_dirty_pertxg[TXG_SIZE];
uint64_t dp_dirty_total;
uint64_t dp_long_free_dirty_pertxg[TXG_SIZE];
uint64_t dp_mos_used_delta;
uint64_t dp_mos_compressed_delta;
uint64_t dp_mos_uncompressed_delta;