ZFS cheatsheet
From Lolly's Wiki
Jump to navigationJump to search
Links
- Wichtige ZFS-Patches: 127729-07 (x86) / 127728-06 (SPARC)
- ZFS Best Practices Guide http://www.solarisinternals.com/wiki/index.php/ZFS_Best_Practices_Guide
- ZFS FAQ bei Opensolaris.org http://www.opensolaris.org/os/community/zfs/faq/
Löschen nicht löschbarer Snapshots
Hier nach einem abgebrochenen ZFS send/recv:
# zfs destroy MYSQL-LOG/binlog@copy_20130403
cannot destroy 'MYSQL-LOG/binlog@copy_20130403': dataset is busy
# zfs holds -r MYSQL-LOG@copy_20130403
NAME TAG TIMESTAMP
MYSQL-LOG@copy_20130403 .send-22887-0 Wed Apr 3 09:03:32 2013
# zfs release .send-22887-0 MYSQL-LOG@copy_20130403
# zfs destroy MYSQL-LOG/binlog@copy_20130403
ZFS Tuning
Eine gefühlte Langsamkeit auf Systemen mit ZFS kommt vom sehr großen Cachehunger. Den kann man eingrenzen: Erstmal schauen, was phase ist:
lollypop@wirefall:~# echo "::kmastat ! grep Total" |mdb -k
Total [hat_memload] 13508608B 309323764 0
Total [kmem_msb] 24010752B 1509706 0
Total [kmem_va] 660340736B 140448 0
Total [kmem_default] 690409472B 1416078794 0
Total [kmem_io_64G] 34619392B 8456 0
Total [kmem_io_4G] 16384B 92 0
Total [kmem_io_2G] 24576B 62 0
Total [bp_map] 1048576B 234488 0
Total [umem_np] 786432B 976 0
Total [id32] 4096B 2620 0
Total [zfs_file_data_buf] 1471275008B 1326646 0
Total [segkp] 589824B 192886 0
Total [ip_minor_arena_sa] 64B 13332 0
Total [ip_minor_arena_la] 192B 45183 0
Total [spdsock] 64B 1 0
Total [namefs_inodes] 64B 24 0
lollypop@wirefall:~# echo "::memstat" | mdb -k
Page Summary Pages MB %Tot
------------ ---------------- ---------------- ----
Kernel 255013 996 24%
ZFS File Data 359196 1403 34%
Anon 346538 1353 33%
Exec and libs 33948 132 3%
Page cache 4836 18 0%
Free (cachelist) 22086 86 2%
Free (freelist) 23420 91 2%
Total 1045037 4082
Physical 1045036 4082
Oder nur ZFS
echo "::memstat ! egrep '(Page Summary|-----|ZFS)'"| mdb -k
Ausgeben aller ARC-Parameter:
lollypop@wirefall:~# echo "::arc -m" | mdb -k
hits = 80839319
misses = 3717788
demand_data_hits = 4127150
demand_data_misses = 51589
demand_metadata_hits = 9467792
demand_metadata_misses = 2125852
prefetch_data_hits = 127941
prefetch_data_misses = 596238
prefetch_metadata_hits = 67116436
prefetch_metadata_misses = 944109
mru_hits = 2031248
mru_ghost_hits = 1906199
mfu_hits = 78514880
mfu_ghost_hits = 993236
deleted = 880714
recycle_miss = 1381210
mutex_miss = 197
evict_skip = 38573528
evict_l2_cached = 0
evict_l2_eligible = 94658370048
evict_l2_ineligible = 8946457600
hash_elements = 79571
hash_elements_max = 82328
hash_collisions = 3005774
hash_chains = 22460
hash_chain_max = 8
p = 64 MB
c = 512 MB
c_min = 127 MB
c_max = 512 MB
size = 512 MB
hdr_size = 14825736
data_size = 468982784
other_size = 53480992
l2_hits = 0
l2_misses = 0
l2_feeds = 0
l2_rw_clash = 0
l2_read_bytes = 0
l2_write_bytes = 0
l2_writes_sent = 0
l2_writes_done = 0
l2_writes_error = 0
l2_writes_hdr_miss = 0
l2_evict_lock_retry = 0
l2_evict_reading = 0
l2_free_on_write = 0
l2_abort_lowmem = 0
l2_cksum_bad = 0
l2_io_error = 0
l2_size = 0
l2_hdr_size = 0
memory_throttle_count = 0
arc_no_grow = 0
arc_tempreserve = 0 MB
arc_meta_used = 150 MB
arc_meta_limit = 128 MB
arc_meta_max = 313 MB
Man kann sich auch alle Parameter ausgeben lassen, die für ZFS gesetzt sind mit:
# echo ::zfs_params | mdb -k
arc_reduce_dnlc_percent = 0x3
zfs_arc_max = 0x100000000
zfs_arc_min = 0x0
arc_shrink_shift = 0x5
zfs_mdcomp_disable = 0x0
zfs_prefetch_disable = 0x0
zfetch_max_streams = 0x8
zfetch_min_sec_reap = 0x2
zfetch_block_cap = 0x100
zfetch_array_rd_sz = 0x100000
zfs_default_bs = 0x9
zfs_default_ibs = 0xe
...
# echo "::arc -a" | mdb -k
hits = 592730
misses = 5095
demand_data_hits = 0
demand_data_misses = 0
demand_metadata_hits = 592719
demand_metadata_misses = 4866
prefetch_data_hits = 0
prefetch_data_misses = 0
...
Setzen von Kernelparametern geht auch online mit:
# echo zfs_arc_max/Z100000000 | mdb -kw
zfs_arc_max: <old value> = 0x100000000
Das setzt den zfs_arc_max auf 4GB = 0x100000000
Limitieren des ARC Cache
In der /etc/system einfach: set zfs:zfs_arc_max = <Number of bytes>
Easy calculation:
# NUMGB=32
# printf "set zfs:zfs_arc_max = 0x%x\n" $[ ${NUMGB} * 1024 ** 3 ]
set zfs:zfs_arc_max = 0x800000000
Siehe auch Limiting the ARC Cache
But !!!! NEVER DO THIS !!!!
Never use mdb -kw to set the values!!!
But on a test system you could try to get the position in the Kernel with
> arc_stats::print -a arcstat_p.value.ui64 arcstat_c.value.ui64 arcstat_c_max.value.ui64
Calculate for example 8GB:
# printf "0x%x\n" $[ 8 * 1024 ** 3 ]
0x200000000
And raise the values like this:
arc.c = arc.c_max arc.p = arc.c / 2
# mdb -kw
Loading modules: [ unix krtld genunix dtrace specfs uppc pcplusmp cpu.generic zfs mpt_sas sockfs ip hook neti dls sctp arp usba uhci fcp fctl qlc nca md lofs sata cpc fcip random crypto logindmux ptm ufs sppp nfs ipc ]
> arc_stats::print -a arcstat_p.value.ui64 arcstat_c.value.ui64 arcstat_c_max.value.ui64
fffffffffbcfaf90 arcstat_p.value.ui64 = 0x4000000
fffffffffbcfafc0 arcstat_c.value.ui64 = 0x40000000
fffffffffbcfb020 arcstat_c_max.value.ui64 = 0x40000000
> fffffffffbcfb020/Z 0x200000000
arc_stats+0x4a0:0x40000000 = 0x200000000
> fffffffffbcfafc0/Z 0x200000000
arc_stats+0x440:0x44a42480 = 0x200000000
> fffffffffbcfaf90/Z 0x100000000
arc_stats+0x410:0x4000000 = 0x100000000
ZFS Platzverbrauch besser anzeigen
$ zfs list -o space NAME AVAIL USED USEDSNAP USEDDS USEDREFRESERV USEDCHILD rpool 25.4G 7.79G 0 64K 0 7.79G rpool/ROOT 25.4G 6.29G 0 18K 0 6.29G rpool/ROOT/snv_98 25.4G 6.29G 0 6.29G 0 0 rpool/dump 25.4G 1.00G 0 1.00G 0 0 rpool/export 25.4G 38K 0 20K 0 18K rpool/export/home 25.4G 18K 0 18K 0 0 rpool/swap 25.8G 512M 0 111M 401M 0
Wenn zfs list -o space als shortcut noch nicht zur Verfügung steht, geht meist:
$ zfs list -o name,avail,used,usedsnap,usedds,usedrefreserv,usedchild -t filesystem,volume
Migration UFS-Root -> ZFS-Root via Live-Upgrade
Erstmal den ZFS Rootpool anlegen:
# zpool create rpool /dev/dsk/<zfs-disk>
Wer Problemen aus dem Weg gehen möchte, läßt den Namen bei rpool.
Boot-Environment (BE) mit lucreate erstellen
# lucreate -c ufsBE -n zfsBE -p rpool
Hiermit werden die Files in die ZFS-Umgebung kopiert.
Prüfen, ob das BootFS richtig gesetzt wurde:
# zpool get bootfs rpool NAME PROPERTY VALUE SOURCE rpool bootfs rpool/ROOT/zfsBE local
Auskommentieren von eventuell noch nachgebliebenen rootdev-Einträgen in der /etc/system
# zpool export rpool # mkdir /tmp/rpool # zpool import -R /tmp/rpool rpool # zfs unmount rpool # rmdir /tmp/rpool/rpool # zfs mount rpool/ROOT/zfsBE # perl -pi.orig -e 's#^(rootdev.*)$#* \1#g' /tmp/rpool/etc/system # zpool export rpool
Bootblock für ZFS auf die ZFS-Platte
# installboot -F zfs /usr/platform/`uname -i`/lib/fs/zfs/bootblk /dev/rdsk/<zfs-disk>
Aktivieren des neuen BEs
# luactivate zfsBE
cannot destroy 'snapshot': dataset is busy
root@sun1 # zfs destroy zpool1/raiddisk0@send_1
cannot destroy 'zpool1/raiddisk0@send_1': dataset is busy
root@sun1 # zfs holds zpool1/raiddisk0@send_1
NAME TAG TIMESTAMP
zpool1/raiddisk0@send_1 .send-14952-0 Mon Jun 15 15:29:09 2015
zpool1/raiddisk0@send_1 .send-16117-0 Mon Jun 15 15:29:28 2015
zpool1/raiddisk0@send_1 .send-26208-0 Tue Jun 16 10:14:47 2015
zpool1/raiddisk0@send_1 .send-8129-0 Mon Jun 15 15:26:54 2015
root@sun1 # zfs release .send-14952-0 zpool1/raiddisk0@send_1
root@sun1 # zfs release .send-16117-0 zpool1/raiddisk0@send_1
root@sun1 # zfs release .send-26208-0 zpool1/raiddisk0@send_1
root@sun1 # zfs release .send-8129-0 zpool1/raiddisk0@send_1
root@sun1 # zfs holds zpool1/raiddisk0@send_1
root@sun1 #
root@sun1 # zfs destroy zpool1/raiddisk0@send_1
root@sun1 #
Fragmentation
# zdb -mm <pool> | nawk '/fragmentation/{count++;frag+=$NF}END{printf "Overall fragmentation %.2d\n",(frag/count);}'
