[DRBD-user] DRBD won't take 1G syncer rate
dbarker at visioncomm.net
Fri Aug 5 20:36:36 CEST 2011
I'm in the middle of a "verify the upgrade procedure" for drbd 8.3.10 to
Just for chucks, I did the dd command here, to see the performance I get.
root at Drbd0:/usr/src/drbd-8.4.0# cat /proc/drbd
version: 8.4.0 (api:1/proto:86-100)
GIT-hash: 28753f559ab51b549d16bcf487fe625d5919c49c build by root at Drbd0,
0: cs:Connected ro:Secondary/Primary ds:UpToDate/UpToDate C r-----
ns:0 nr:0 dw:0 dr:0 al:0 bm:0 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:0
root at Drbd0:/usr/src/drbd-8.4.0# drbdadm primary r0
root at Drbd0:/usr/src/drbd-8.4.0# dd if=/dev/zero of=/dev/drbd0 bs=1M
dd: writing `/dev/drbd0': No space left on device
5120+0 records in
5119+0 records out
5368508416 bytes (5.4 GB) copied, 499.712 s, 10.7 MB/s
Now, 11MB/s doesn't sound like much, but you have to remember:
a) These are drbd virtual machines whose disks are connected (so the
I/O has to go through the wire to the other testing drbd in a different ESXi
b) These target disks are actually virtual disks,
c) Those virtual disks are on the "real" drbd machines (not yet
upgraded, actually, they are virtual machines in the same 2 ESXi hosts)
d) The real drbd disks are using the same 1G nic to communicate.
So, your path was dd => raid card => 7.2Krpm disk = poor performance.
My path is dd => connected drbd device => vmfs filesystem => iSCSI initiator
=> iSCSI Target (virtual) => connected drbd device => vmfs filesystem =>
MoBo controller => 7.2Krpm sata disk = the same performance.
I'm pleased with my poor performance<g>.
BTW, there wasn't much in the "procedure" just:
tar -xzf drbd-8.4.0.tar.gz
./configure --with-km --prefix /usr --sysconfdir /etc --localstatedir /var
make clean all
drbdadm dump, convert configs to new format
Worked a charm!
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the drbd-user