Mailing List Archive

* Raul Oliveira <rto@example.com> [010926 11:02]:
> Hi
>
> I finally got the cluster working in a more stable condition with kernel
> 2.4.9 and drbd-0.6.1-pre2. With the kernel 2.2.18 and 0.5.8.1 I used to
> have one machine blocking completly, not completly because I could ping it!
>
> Anyway now it seems much better.
> My problem is that I have 3 disks of 37GB in each machine, and when there
> is full-sync it takes lots of time. So I need to use the sync-rate as fast
> as I can.
> Each server has 2 network cards, and I'm using the one of this cards (eth1)
> with a crossover cable, to pass hearbeat "pings" and drbd disk
> syncronization. Is there any problem if I use as sync-rate for both of my
> two resources (drbd0 and drbd1) 5000KB. This means that I will be nearly
> over the network capacity (100Mbps) if both resources need syncronisation.
> Shall I use for example 4000KB.
> In order to introduce some flexibility between my two resources, I'm trying
> to configure LVM to define the filesystems associated with drbd resources.
> Is there any problem?
>
> Best regards
> Raul
>

I can not recomend you to allocate more that the available bandwith
for synchronisation.

What about a second pair of NICs ?

LVM should work, never tested. (Maybe we will need entries in
/proc/partitions), please share your experiences with us.

PS: You can try a sync rate up to 6MB/s. (I have never tested if it really
reaches that speed, it could work in theroy)

-Philipp

Hi Philipp

> I can not recomend you to allocate more that the available bandwith
> for synchronisation.

Ok, I asked because I didn't know how you use this sync-rate.

> What about a second pair of NICs ?

I'm already using a second pair of NICs for heartbeat and drbd
synchronization. In each system I have eth0 for servers address and cluster
address, eth1 connecting both servers with a crossover cable.


> LVM should work, never tested. (Maybe we will need entries in
> /proc/partitions), please share your experiences with us.

It works, but i have the following result for drbesetup:

ServerL211:/home/operador# drbdsetup /dev/nb0 show
Lower device: 58:00 ((null))
Disk options:
Local address: 10.0.0.2:7788
Remote address: 10.0.0.1:7788
Wire protocol: C
Net options:
timeout = 6.0 sec
sync-rate = 5000 KB/sec
tl-size = 256
connect-int = 10 sec
ping-int = 10 sec

ServerL211:/home/operador# drbdsetup /dev/nb1 show
Lower device: 58:01 ((null))
Disk options:
Local address: 10.0.0.2:7789
Remote address: 10.0.0.1:7789
Wire protocol: C
Net options:
timeout = 6.0 sec
sync-rate = 5000 KB/sec
tl-size = 256
connect-int = 10 sec
ping-int = 10 sec

resource drbd0 {

protocol=C
fsckcmd=fsck -p -y

net {
sync-rate=5000
tl-size=256
timeout=60
connect-int=10
ping-int=10
}

on ServerL210 {
device=/dev/nb0
disk=/dev/dados/contas
address=10.0.0.1
port=7788
}

on ServerL211 {
device=/dev/nb0
disk=/dev/dados/contas
address=10.0.0.2
port=7788
}
}

resource drbd1 {

protocol=C
fsckcmd=fsck -p -y

net {
sync-rate=5000
tl-size=256
timeout=60
connect-int=10
ping-int=10
}

on ServerL210 {
device=/dev/nb1
disk=/dev/dados/gestao
address=10.0.0.1
port=7789
}

on ServerL211 {
device=/dev/nb1
disk=/dev/dados/gestao
address=10.0.0.2
port=7789
}
}

> PS: You can try a sync rate up to 6MB/s. (I have never tested if it
> really
> reaches that speed, it could work in theroy)

Each time I use two drbd ressources I got very low speeds in writing in the
partitions being mirrored by drbd. I mean very very slow. And sometimes
the slave server blocks (no response art all).

Best regards
Raul

Hi

> Each time I use two drbd ressources I got very low speeds in writing in
> the
> partitions being mirrored by drbd. I mean very very slow. And sometimes
>
> the slave server blocks (no response art all).

I was forgetting to create the filesystems after the initialization of /dev/nb?
with the drbdsetup.
I just performed a mkfs and everything gots clean. Since these filesystems were
created before to test LVM, I didn't make it again. It was funny because the
system was working but very very slow. And sometime the slave blocked!!!
I don't know if this stupid experiences are usefull for you, but in anycase.
Thanks
Raul

* rto@example.com <rto@fe.up.pt> [010926 23:58]:
> Hi Philipp
>
> > I can not recomend you to allocate more that the available bandwith
> > for synchronisation.
>
> Ok, I asked because I didn't know how you use this sync-rate.
>
> > What about a second pair of NICs ?
>
> I'm already using a second pair of NICs for heartbeat and drbd
> synchronization. In each system I have eth0 for servers address and cluster
> address, eth1 connecting both servers with a crossover cable.
>

Ok, then I mean a third pair of NICs ?
(A second pair for DRBD)
But with todays max sync rate of 6MB/sec this is not a big gain.
As soon as DRBD can offer a sync rate of 8 or 9 it will make sense.

>
> > LVM should work, never tested. (Maybe we will need entries in
> > /proc/partitions), please share your experiences with us.
>
> It works, but i have the following result for drbesetup:
>
> ServerL211:/home/operador# drbdsetup /dev/nb0 show
> Lower device: 58:00 ((null))
^^^^^^^^
This is only a cosmetic error. drbdsetup searches under /dev/ vor the
name of the device. If it encounters an error, or can not find the
device node for 58:00 it should print "can not guess device name".
There seems to be a problem, since it prints (null)

> Disk options:
> Local address: 10.0.0.2:7788
> Remote address: 10.0.0.1:7788
> Wire protocol: C
> Net options:
> timeout = 6.0 sec
> sync-rate = 5000 KB/sec
> tl-size = 256
> connect-int = 10 sec
> ping-int = 10 sec
>
> ServerL211:/home/operador# drbdsetup /dev/nb1 show
> Lower device: 58:01 ((null))
> Disk options:
> Local address: 10.0.0.2:7789
> Remote address: 10.0.0.1:7789
> Wire protocol: C
> Net options:
> timeout = 6.0 sec
> sync-rate = 5000 KB/sec
> tl-size = 256
> connect-int = 10 sec
> ping-int = 10 sec
>
> resource drbd0 {
>
> protocol=C
> fsckcmd=fsck -p -y
>
> net {
> sync-rate=5000
> tl-size=256
> timeout=60
> connect-int=10
> ping-int=10
> }
>
> on ServerL210 {
> device=/dev/nb0
> disk=/dev/dados/contas
> address=10.0.0.1
> port=7788
> }
>
> on ServerL211 {
> device=/dev/nb0
> disk=/dev/dados/contas
> address=10.0.0.2
> port=7788
> }
> }
>
> resource drbd1 {
>
> protocol=C
> fsckcmd=fsck -p -y
>
> net {
> sync-rate=5000
> tl-size=256
> timeout=60
> connect-int=10
> ping-int=10
> }
>
> on ServerL210 {
> device=/dev/nb1
> disk=/dev/dados/gestao
> address=10.0.0.1
> port=7789
> }
>
> on ServerL211 {
> device=/dev/nb1
> disk=/dev/dados/gestao
> address=10.0.0.2
> port=7789
> }
> }
>
> > PS: You can try a sync rate up to 6MB/s. (I have never tested if it
> > really
> > reaches that speed, it could work in theroy)
>
> Each time I use two drbd ressources I got very low speeds in writing in the
> partitions being mirrored by drbd. I mean very very slow. And sometimes
> the slave server blocks (no response art all).
>
> Best regards
> Raul
>
>
> _______________________________________________
> DRBD-devel mailing list
> DRBD-devel@example.com
> https://lists.sourceforge.net/lists/listinfo/drbd-devel