[DRBD-user] the perfomance issuce with a physical volume synchronization

Junko IKEDA tsukishima.ha at gmail.com
Fri May 20 21:22:51 CEST 2011

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


Hi,

> DRBD version, LVM version, Device mapper version (kernel version),
> distribution?

DRBD version
8.3.10(we noticed this with 8.3.5 at first, and update DRBD after that)

LVM version
LVM2(included RHEL in 5.2)

kernel version
2.6.18-92.EL5 (x86_64)

distribution
RHEL 5.2

> What about oflag=dsync instead of direct?
> What about not specifying any oflag, but doing
> # dd if=/dev/zero of=/mnt/dd/10GB.dat bs=1M count=10000 conv=notrunc,fsync

OK, I'll try it.

> What about reads?
> With iflag=direct?

We just felt that reads were also slow,
but we didn't record them as mathematical values.

> Without without iflag, but still cache cold,
> after an echo 3 > /proc/sys/vm/drop_caches?

yes, we went clear the drop_caches,
and also did "mount/unmount" the device.

> Do you have an other data point with more recent (all of it),
> especially with more recent LVM + upstream kernel?

We tried the same test on RHEL 5.2、5.5、6.1,
but  the results was the same.

Thanks,
Junko IKEDA



More information about the drbd-user mailing list