Bonjour,
Alors voici les tests d'une synchro de 4To sur l'interface Giga des deux serveurs :
root@node1:/etc/drbd.d# drbdsetup /dev/drbd0 syncer -r 1000M
root@node1:/etc/drbd.d# cat /proc/drbd
version: 8.3.11 (api:88/proto:86-96)
srcversion: F937DCB2E5D83C6CCE4A6C9
0: cs:SyncTarget ro:Primary/Secondary ds:Inconsistent/UpToDate C r-----
ns:0 nr:9537420 dw:9537408 dr:8 al:0 bm:582 lo:1 pe:7410 ua:0 ap:1 ep:1 wo:f oos:3897360024
[>....................] sync'ed: 0.3% (3806012/3815328)Mfinish: 14:40:25 speed: 73,756 (70,644) want: 1,000,001 K/sec
#### atop
PRC | sys 2.41s | user 0.00s | | | #proc 72 | | #trun 1 | #tslpi 73 | #tslpu 1 | | #zombie 0 | clones 0 | | | #exit 0 |
CPU | sys 19% | user 0% | | irq 1% | | | idle 275% | wait 105% | | | steal 0% | guest 0% | | curf 2.80GHz | curscal ?% |
cpu | sys 19% | user 0% | | irq 1% | | | idle 80% | cpu003 w 0% | | | steal 0% | guest 0% | | curf 2.80GHz | curscal ?% |
cpu | sys 2% | user 0% | | irq 0% | | | idle 98% | cpu001 w 0% | | | steal 0% | guest 0% | | curf 2.80GHz | curscal ?% |
cpu | sys 0% | user 0% | | irq 0% | | | idle 100% | cpu002 w 0% | | | steal 0% | guest 0% | | curf 2.80GHz | curscal ?% |
cpu | sys 0% | user 0% | | irq 0% | | | idle 0% | cpu000 w100% | | | steal 0% | guest 0% | | curf 2.80GHz | curscal ?% |
CPL | avg1 1.95 | | avg5 0.84 | | avg15 0.78 | | | | csw 108392 | | intr 53239 | | | | numcpu 4 |
MEM | tot 2.0G | | free 1.7G | cache 71.0M | | dirty 0.5M | buff 4.7M | | slab 17.9M | | | | | | |
SWP | tot 4.0G | | free 4.0G | | | | | | | | | | vmcom 41.0M | | vmlim 5.0G |
DSK | sdb | | busy 70% | read 0 | | write 5323 | KiB/r 0 | | KiB/w 131 | MBr/s 0.00 | | MBw/s 68.59 | avq 1.27 | | avio 1.25 ms |
DSK | sda | | busy 1% | read 4 | | write 2 | KiB/r 7 | | KiB/w 38 | MBr/s 0.00 | | MBw/s 0.01 | avq 1.57 | | avio 9.33 ms |
NET | transport | tcpi 47361 | | tcpo 46753 | udpi 0 | udpo 0 | tcpao 0 | | tcppo 0 | tcprs 0 | tcpie 0 | tcpor 0 | | udpnp 0 | udpip 0 |
NET | network | | ipi 47363 | ipo 46754 | | ipfrw 0 | deliv 47363 | | | | | | icmpi 0 | | icmpo 0 |
NET | eth1 60% | pcki 502072 | | pcko 46743 | si 601 Mbps | | so 2749 Kbps | coll 0 | mlti 0 | | erri 0 | erro 0 | | drpi 0 | drpo 0 |
NET | eth0 0% | pcki 17 | | pcko 11 | si 1 Kbps | | so 10 Kbps | coll 0 | mlti 0 | | erri 0 | erro 0 | | drpi 0 | drpo 0 |
PID RUID EUID THR SYSCPU USRCPU VGROW RGROW RDDSK WRDSK ST EXC S CPUNR CPU CMD 1/1
2575 root root 1 1.94s 0.00s 0K 0K 0K 0K -- - S 3 20% drbd0_receiver
2580 root root 1 0.37s 0.00s 0K 0K 0K 0K -- - S 3 4% drbd0_asender
2560 root root 1 0.04s 0.00s 0K 0K 0K 0K -- - S 1 0% drbd0_worker
2860 root root 1 0.02s 0.00s 8888K 728K 0K 0K -- - R 2 0% atop
2512 root root 1 0.02s 0.00s 0K 0K 0K 0K -- - S 3 0% kworker/3:0
2533 root root 1 0.02s 0.00s 0K 0K 0K 0K -- - S 3 0% kworker/3:1
2096 root root 1 0.00s 0.00s 0K 0K 0K 0K -- - S 3 0% sshd
1915 root root 1 0.00s 0.00s 28K 472K 28K 0K -- - S 1 0% cron
25 root root 1 0.00s 0.00s 0K 0K 0K 0K -- - S 2 0% sync_supers
176 root root 1 0.00s 0.00s 0K 0K 0K 4K -- - S 3 0% jbd2/sda1-8
2531 root root 1 0.00s 0.00s 0K 0K 0K 0K -- - S 2 0% flush-8:0
### iftop -i eth1
node1 => node2 2,16Mb 2,15Mb 2,14Mb
<= 550Mb 549Mb 549Mb
Date: Sun, 30 Mar 2014 00:57:18 +1100
From: valferon@gmail.com
To: frsag@frsag.org
Subject: Re: [FRsAG] DRBD valeur syncer
Hello,
En passant le syncer a 1000M au lieu de 100M, la difference est minime etant donne qu'en admettant que ton lien est le bottleneck et pas les disques, la vitesse theorique maximale est de 125MB/s.
A 100M/s tu es donc pas loin de ta vitesse maximum de synchro.
Comme le dit Olivier, 30% une fois la premiere synchro faite est recommande et un bon compromis.
J'aime bien 'atop' personellement pour mettre en evidence ce genre de limitations (hdd, cpu, mem, inet}.
My 2 cents.
Val
_______________________________________________
Liste de diffusion du FRsAG
http://www.frsag.org/