Date: Fri, 01 Aug 2014 15:35:49 -0400 From: Mike Tancsa <mike@sentex.net> To: Mikolaj Golub <to.my.trociny@gmail.com> Cc: freebsd-fs@freebsd.org Subject: Re: HAST slowness ? Message-ID: <53DBEC15.2040406@sentex.net> In-Reply-To: <20140801183005.GA3286@gmail.com> References: <53DAB322.1050703@sentex.net> <20140801183005.GA3286@gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 8/1/2014 2:30 PM, Mikolaj Golub wrote:
> On Thu, Jul 31, 2014 at 05:20:34PM -0400, Mike Tancsa wrote:
>> I am trying to setup my first HAST setup and for some reason, traffic
>> sync seems to be really slow. If I do something simple like
>>
>>
>> #dd if=/dev/zero of=/hastvirt/test count=2000 bs=1024k
>>
>> on the primary box, traffic on the secondary never gets much beyond a
>> Mb/s ?!?!
>>
>
> It looks you are writing to a file-system, not directly to HAST
> device? What is this FS? I see two HAST devices? Are they combined in
> one FS? How? What if you write directly to a HAST device instead?
Hi Mikolaj,
Thank you for responding. I have 2 servers, both with two disks. Each
have 2 zfs partitions The first zfs partition is the operating system
ada0p4 and ada1p4 are for hast on each of the boxes
# zpool status
pool: hastvirt
state: ONLINE
scan: scrub repaired 0 in 0h5m with 0 errors on Thu Jul 31 17:34:49 2014
config:
NAME STATE READ WRITE CKSUM
hastvirt ONLINE 0 0 0
mirror-0 ONLINE 0 0 0
hast/disk1 ONLINE 0 0 0
hast/disk2 ONLINE 0 0 0
The idea if one box dies, the other can take over and mount hastvirt
>
> Also note, "hole" compression is enabled by default, so your test with
> copying zero blocks is not very good. You might be limited here by
> RTT, because in memsync mode, on every "compressed" block sent HAST
> waits for response before proceeding with the next block.
Perhaps this indeed was the issue. I tried importing a windows VM and it
seemed to go much faster. I am now seeing traffic across the HAST
dedicated NIC at close to 200Mb
% ifstat -b -i vlan5
vlan5
Kbps in Kbps out
262365.4 6407.79
235609.5 5757.97
217813.4 5380.81
144054.8 3606.39
150899.8 3882.17
240806.7 5908.37
238943.7 5718.61
Looking at gstat on the master, it currently has maxed out the disk
writing so I guess the network is keeping up.
# hastctl status
Name Status Role Components
disk1 complete primary /dev/ada0p4 1.1.1.1
disk2 complete primary /dev/ada1p4 1.1.1.1
hastctl list
disk1:
role: primary
provname: disk1
localpath: /dev/ada0p4
extentsize: 2097152 (2.0MB)
keepdirty: 64
remoteaddr: 1.1.1.1
replication: memsync
status: complete
workerpid: 1021
dirty: 4194304 (4.0MB)
statistics:
reads: 387530
writes: 309272
deletes: 11
flushes: 480
activemap updates: 19121
local errors: read: 0, write: 0, delete: 11, flush: 0
queues: local: 0, send: 0, recv: 0, done: 0, idle: 255
disk2:
role: primary
provname: disk2
localpath: /dev/ada1p4
extentsize: 2097152 (2.0MB)
keepdirty: 64
remoteaddr: 1.1.1.1
replication: memsync
status: complete
workerpid: 1023
dirty: 4194304 (4.0MB)
statistics:
reads: 387008
writes: 309251
deletes: 11
flushes: 480
activemap updates: 19118
local errors: read: 0, write: 0, delete: 11, flush: 0
queues: local: 0, send: 0, recv: 0, done: 0, idle: 255
Are there any good hast tuning guides beyond the wiki you recommend ?
Should anything be tuned beyond the defaults ?
Thanks again
---Mike
--
-------------------
Mike Tancsa, tel +1 519 651 3400
Sentex Communications, mike@sentex.net
Providing Internet services since 1994 www.sentex.net
Cambridge, Ontario Canada http://www.tancsa.com/
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?53DBEC15.2040406>