Hi,
I have a HAST device set up between two systems. I experience very low
speed with dirty blocks synchronization after split-brain condition
been recovered: it's 200KB/s average on 1Gbit link. On the other side,
when i copy a big file to the zfs partition that is created on top of
the hast device the synchronization speed between the host is 50MB/s
(wich is not too high for 1Gbit link, but acceptable.)
uname -a
FreeBSD rest 8.2-STABLE FreeBSD 8.2-STABLE #3: Tue May 31 18:51:19
EEST 2011 root@rest:/usr/obj/usr/src/sys/GENERIC amd64
Both systems have the same hardware and FreeBSD version.
cat /etc/hast.conf
resource mirror0 {
local /dev/mfid0s3
on rest2 {
remote 192.168.1.51
}
on rest {
remote 192.168.1.52
}
}
hastctl status
mirror0:
role: primary
provname: mirror0
localpath: /dev/mfid0s3
extentsize: 2097152 (2.0MB)
keepdirty: 64
remoteaddr: 192.168.3.53
replication: fullsync
status: complete
dirty: 26944208896 (25GB)
df -h
Filesystem Size Used Avail Capacity Mounted on
failover 34G 31G 2.9G 91% /failover
The /dev/mfid0s3 size is 34G, and synchronization has been running for
19 hours and still 25G remaining. That's terribly slow! What could be
the reason of that?
The other question that bothers me is that i filled the partition to
91% and it was written to the secondary host (at list the data was
transfered to the secondary host at rate 50MB/s), but hast still
showed all 34G of data as dirty after the copying. Why is that so?
P.S. I am not in the list, so please, CC