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>
This commit is contained in:
parent
aed3e94d2f
commit
1e37d7e555
Notes:
svn2git
2020-12-20 02:59:44 +00:00
svn path=/vendor-sys/illumos/dist/; revision=312436
@ -56,6 +56,14 @@
|
||||
*/
|
||||
int zfs_nopwrite_enabled = 1;
|
||||
|
||||
/*
|
||||
* 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;
|
||||
|
||||
const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = {
|
||||
{ DMU_BSWAP_UINT8, TRUE, "unallocated" },
|
||||
{ DMU_BSWAP_ZAP, TRUE, "object directory" },
|
||||
@ -717,15 +725,25 @@ 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;
|
||||
|
||||
if (dmu_objset_zfs_unmounting(dn->dn_objset))
|
||||
return (SET_ERROR(EINTR));
|
||||
@ -739,9 +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
|
||||
@ -753,10 +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);
|
||||
}
|
||||
|
@ -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>
|
||||
@ -492,6 +493,16 @@ 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.
|
||||
|
@ -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;
|
||||
|
Loading…
Reference in New Issue
Block a user