Hi.
The problem is getting worse... now even if I destroy all snapshots in a pool
I get performance problem even with zil_disable set to 1.
Despite that I have limit for maximum nfs threads set to 2048 I get only about
1700.
If I want to kill nfsd server it takes 1-4 minutes untill all threads are
finished (and most of the time above 1000 threads are stil there). During nfsd
is stopping I can see IOs (99% reads) to the pool.
Also simple zfs commands (like changing quote for a file system, etc.) take too
much time to complete (like 3-5 minutes to set quota for a file system).
There's still over 700gb free storage in a pool, setting quota to none doesn't
help.
Using iostat it looks like disks aren't saturated. It looks almost like lot of
nfsd threads are spinning (probably in zfs) - I don't recall nfsd stopping so
long on UFS file systems.
bash-3.00# iostat -xnzC 1
[... first output]
extended device statistics
r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
1782.3 0.0 12700.7 0.0 0.0 11.1 0.0 6.2 0 293 c5
584.3 0.0 4102.2 0.0 0.0 3.6 0.0 6.2 0 96
c5t600C0FF000000000098FD57F9DA83C00d0
572.1 0.0 4144.8 0.0 0.0 3.7 0.0 6.5 0 99
c5t600C0FF000000000098FD55DBA4EA000d0
625.8 0.0 4453.7 0.0 0.0 3.7 0.0 5.9 0 98
c5t600C0FF000000000098FD516E4403200d0
extended device statistics
r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
2381.9 0.0 17242.1 0.0 0.0 16.7 0.0 7.0 0 291 c5
800.0 0.0 5827.9 0.0 0.0 5.5 0.0 6.9 0 96
c5t600C0FF000000000098FD57F9DA83C00d0
690.0 0.0 4991.9 0.0 0.0 4.9 0.0 7.1 0 97
c5t600C0FF000000000098FD55DBA4EA000d0
892.0 0.0 6422.3 0.0 0.0 6.3 0.0 7.1 0 98
c5t600C0FF000000000098FD516E4403200d0
extended device statistics
r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
1890.2 0.0 12826.0 0.0 0.0 11.4 0.0 6.0 0 292 c5
611.1 0.0 3998.5 0.0 0.0 3.4 0.0 5.6 0 96
c5t600C0FF000000000098FD57F9DA83C00d0
604.1 0.0 4259.5 0.0 0.0 3.9 0.0 6.4 0 98
c5t600C0FF000000000098FD55DBA4EA000d0
675.1 0.0 4568.0 0.0 0.0 4.1 0.0 6.0 0 98
c5t600C0FF000000000098FD516E4403200d0
^C
Now dtrace output you've requested.
dtrace -n '[EMAIL PROTECTED](20)] = count()}' -c 'sleep 5'
[...]
unix`i_ddi_splhigh
unix`disp_getwork+0x38
unix`idle+0xd4
unix`thread_start+0x4
35
unix`i_ddi_splx
unix`disp_getwork+0x160
unix`idle+0xd4
unix`thread_start+0x4
38
unix`disp_getwork+0x70
unix`idle+0xd4
unix`thread_start+0x4
38
unix`disp_getwork+0x158
unix`idle+0xd4
unix`thread_start+0x4
38
unix`i_ddi_splhigh+0x14
unix`disp_getwork+0x38
unix`idle+0xd4
unix`thread_start+0x4
39
unix`disp_getwork+0x8c
unix`idle+0xd4
unix`thread_start+0x4
39
unix`idle+0x12c
unix`thread_start+0x4
44
unix`disp_getwork+0x1a8
unix`idle+0xd4
unix`thread_start+0x4
47
unix`disp_getwork+0x7c
unix`idle+0xd4
unix`thread_start+0x4
49
unix`disp_getwork+0x90
unix`idle+0xd4
unix`thread_start+0x4
56
unix`disp_getwork+0x10c
unix`idle+0xd4
unix`thread_start+0x4
62
unix`i_ddi_splx+0x1c
unix`disp_getwork+0x160
unix`idle+0xd4
unix`thread_start+0x4
117
bash-3.00#
This message posted from opensolaris.org
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
