* syncing RAID1 with more than 10MB/sec
@ 2005-03-30 9:12 Nils-Henner Krueger
2005-03-30 9:15 ` Catalin(ux aka Dino) BOIE
2005-03-30 9:40 ` Guy
0 siblings, 2 replies; 4+ messages in thread
From: Nils-Henner Krueger @ 2005-03-30 9:12 UTC (permalink / raw)
To: linux-raid
Syncing a raid1 takes long time for large disks because of
the 10MB/sec limit:
Mar 30 10:58:31 imap1 kernel: raid1: raid set md2 not clean;
reconstructing mirrors
Mar 30 10:58:31 imap1 kernel: raid1: raid set md2 active with 2 out of 2
mirrors
Mar 30 10:58:31 imap1 kernel: md: syncing RAID array md2
Mar 30 10:58:31 imap1 kernel: md: minimum _guaranteed_ reconstruction
speed: 100 KB/sec/disc.
Mar 30 10:58:31 imap1 kernel: md: using maximum available idle IO
bandwith (but not more than 10000 KB/sec) for reconstruction.
Is it possible to raise this value (without recompiling)?
I know about possible performance tradeoffs with other
processes, I'm just generally interested.
Thanks.
nils-henner
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: syncing RAID1 with more than 10MB/sec
2005-03-30 9:12 syncing RAID1 with more than 10MB/sec Nils-Henner Krueger
@ 2005-03-30 9:15 ` Catalin(ux aka Dino) BOIE
2005-03-30 9:40 ` Guy
1 sibling, 0 replies; 4+ messages in thread
From: Catalin(ux aka Dino) BOIE @ 2005-03-30 9:15 UTC (permalink / raw)
To: Nils-Henner Krueger; +Cc: linux-raid
> Is it possible to raise this value (without recompiling)?
> I know about possible performance tradeoffs with other
> processes, I'm just generally interested.
>
> Thanks.
echo xxx > /proc/sys/dev/raid/speed_limit_max
---
Catalin(ux aka Dino) BOIE
catab at deuroconsult.ro
http://kernel.umbrella.ro/
^ permalink raw reply [flat|nested] 4+ messages in thread
* RE: syncing RAID1 with more than 10MB/sec
2005-03-30 9:12 syncing RAID1 with more than 10MB/sec Nils-Henner Krueger
2005-03-30 9:15 ` Catalin(ux aka Dino) BOIE
@ 2005-03-30 9:40 ` Guy
2005-03-31 6:25 ` Tim Moore
1 sibling, 1 reply; 4+ messages in thread
From: Guy @ 2005-03-30 9:40 UTC (permalink / raw)
To: 'Nils-Henner Krueger', linux-raid
> -----Original Message-----
> From: linux-raid-owner@vger.kernel.org [mailto:linux-raid-
> owner@vger.kernel.org] On Behalf Of Nils-Henner Krueger
> Sent: Wednesday, March 30, 2005 4:13 AM
> To: linux-raid@vger.kernel.org
> Subject: syncing RAID1 with more than 10MB/sec
>
> Syncing a raid1 takes long time for large disks because of
> the 10MB/sec limit:
>
> Mar 30 10:58:31 imap1 kernel: raid1: raid set md2 not clean;
> reconstructing mirrors
> Mar 30 10:58:31 imap1 kernel: raid1: raid set md2 active with 2 out of 2
> mirrors
> Mar 30 10:58:31 imap1 kernel: md: syncing RAID array md2
> Mar 30 10:58:31 imap1 kernel: md: minimum _guaranteed_ reconstruction
> speed: 100 KB/sec/disc.
> Mar 30 10:58:31 imap1 kernel: md: using maximum available idle IO
> bandwith (but not more than 10000 KB/sec) for reconstruction.
>
> Is it possible to raise this value (without recompiling)?
> I know about possible performance tradeoffs with other
> processes, I'm just generally interested.
>
> Thanks.
>
> nils-henner
Yes. For details:
man md
The default of 10,000 has been increased to 100,000 in newer kernels.
I don't know which kernels.
If you want to override the defaults you could add these 3 lines to
/etc/sysctl.conf:
# RAID rebuild min/max speed K/Sec per device
dev.raid.speed_limit_min = 1000
dev.raid.speed_limit_max = 100000
Adjust above as required. The above will take effect after a reboot.
To see current values type these 2 lines:
cat /proc/sys/dev/raid/speed_limit_min
cat /proc/sys/dev/raid/speed_limit_max
To temporarily change the defaults use these 2 commands:
echo 1000 > /proc/sys/dev/raid/speed_limit_min
echo 100000 > /proc/sys/dev/raid/speed_limit_max
Adjust above as required. The above will take effect now.
Guy
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: syncing RAID1 with more than 10MB/sec
2005-03-30 9:40 ` Guy
@ 2005-03-31 6:25 ` Tim Moore
0 siblings, 0 replies; 4+ messages in thread
From: Tim Moore @ 2005-03-31 6:25 UTC (permalink / raw)
To: linux-raid
Generally the min/max values make no difference as long as max is > 45000.
System use during resync is the biggest factor other than raw CPU.
Check actual resync times:
> cat /var/log/messages.8 | synctime
array, blocks, sec, blocks/s, start, finish
md10, 2104448, 63, 33403.9, Mar 16 23:33:37, Mar 16 23:34:40
md11, 2096384, 64, 32756.0, Mar 16 23:35:30, Mar 16 23:36:34
md13, 2096384, 127, 16507.0, Mar 16 23:36:34, Mar 16 23:37:37
md12, 2096384, 191, 10975.8, Mar 16 23:37:37, Mar 16 23:38:41
md12, 2104448, 61, 34499.1, Mar 16 23:50:38, Mar 16 23:51:39
md5, 2096384, 63, 33275.9, Mar 16 23:51:50, Mar 16 23:52:53
md12, 2104448, 61, 34499.1, Mar 17 00:00:39, Mar 17 00:01:40
md6, 2104384, 122, 17249.0, Mar 17 00:01:40, Mar 17 00:02:41
md5, 2104384, 183, 11499.4, Mar 17 00:02:41, Mar 17 00:03:42
md7, 6289344, 366, 17184.0, Mar 17 00:03:42, Mar 17 00:06:45
md9, 52170944, 1873, 27854.2, Mar 17 00:06:45, Mar 17 00:31:52
md8, 52187008, 3489, 14957.6, Mar 17 00:31:52, Mar 17 00:58:48
> cat ~/bin/synctime | cksum
1035800810 528
> cat ~/bin/synctime
#!/bin/awk -f
#GNU Awk 3.1.1
#Copyright (C) 1989, 1991-2002 Free Software Foundation.
#md resync time csv
BEGIN{
printf "array, blocks, sec, blocks/s, start, finish\n"
}
/md: syncing RAID array/{
sdate = $1 " " $2 " " $3
"date -d \"" sdate "\" +%s" | getline start
dev = $10
}
/md: using.*window,/{
size = $14
}
/md:.*sync done\./{
fdate = $1 " " $2 " " $3
"date -d \"" fdate "\" +%s" | getline finish
delta = finish-start
printf "%4s, %10s, %5s, %6.1f, %s, %s\n",
dev,size,delta,size/delta,sdate,fdate
}
#end
Guy wrote:
>>-----Original Message-----
>>From: linux-raid-owner@vger.kernel.org [mailto:linux-raid-
>>owner@vger.kernel.org] On Behalf Of Nils-Henner Krueger
>>Sent: Wednesday, March 30, 2005 4:13 AM
>>To: linux-raid@vger.kernel.org
>>Subject: syncing RAID1 with more than 10MB/sec
>>
>>Syncing a raid1 takes long time for large disks because of
>>the 10MB/sec limit:
>>
>>Mar 30 10:58:31 imap1 kernel: raid1: raid set md2 not clean;
>>reconstructing mirrors
>>Mar 30 10:58:31 imap1 kernel: raid1: raid set md2 active with 2 out of 2
>>mirrors
>>Mar 30 10:58:31 imap1 kernel: md: syncing RAID array md2
>>Mar 30 10:58:31 imap1 kernel: md: minimum _guaranteed_ reconstruction
>>speed: 100 KB/sec/disc.
>>Mar 30 10:58:31 imap1 kernel: md: using maximum available idle IO
>>bandwith (but not more than 10000 KB/sec) for reconstruction.
>>
>>Is it possible to raise this value (without recompiling)?
>>I know about possible performance tradeoffs with other
>>processes, I'm just generally interested.
>>
>>Thanks.
>>
>>nils-henner
>
>
> Yes. For details:
> man md
>
> The default of 10,000 has been increased to 100,000 in newer kernels.
> I don't know which kernels.
>
> If you want to override the defaults you could add these 3 lines to
> /etc/sysctl.conf:
> # RAID rebuild min/max speed K/Sec per device
> dev.raid.speed_limit_min = 1000
> dev.raid.speed_limit_max = 100000
>
> Adjust above as required. The above will take effect after a reboot.
>
> To see current values type these 2 lines:
> cat /proc/sys/dev/raid/speed_limit_min
> cat /proc/sys/dev/raid/speed_limit_max
>
> To temporarily change the defaults use these 2 commands:
> echo 1000 > /proc/sys/dev/raid/speed_limit_min
> echo 100000 > /proc/sys/dev/raid/speed_limit_max
>
> Adjust above as required. The above will take effect now.
>
> Guy
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
--
| for direct mail add "private_" in front of user name
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2005-03-31 6:25 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-03-30 9:12 syncing RAID1 with more than 10MB/sec Nils-Henner Krueger
2005-03-30 9:15 ` Catalin(ux aka Dino) BOIE
2005-03-30 9:40 ` Guy
2005-03-31 6:25 ` Tim Moore
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).