Page 3 of 4

Re: HA iSCSI Target with DRBD (2 node cluster how-to)

Posted: 2019/08/10 15:32:03
by TrevorH
If it's the same as DRBD8 then you need to bump up:

Code: Select all

disk { al-extents 3389; }
net {                 max-buffers 72000;
                max-epoch-size 20000;
                sndbuf-size 1M;
}

Re: HA iSCSI Target with DRBD (2 node cluster how-to)

Posted: 2019/08/10 16:39:42
by hunter86_bg
Post ping results of '-c 4' .
Are you using one 10GbE interface or multiple small ones ?
You can check with 'iftop' current network usage.

Re: HA iSCSI Target with DRBD (2 node cluster how-to)

Posted: 2019/08/10 16:45:49
by hunter86_bg
Have you checked this one ?

Re: HA iSCSI Target with DRBD (2 node cluster how-to)

Posted: 2019/08/10 20:42:40
by ladam@ictuniverse.eu
So,

It's one 10 GB interface on each server directly connected via direct fiber. Only used for sync. MTU set to 9000

[root@san3 ~]# ethtool ens3f4 | grep -i speed
Speed: 10000Mb/s
[root@san3 ~]#


[root@san3 ~]# ping 10.0.0.31 -c 4
PING 10.0.0.31 (10.0.0.31) 56(84) bytes of data.
64 bytes from 10.0.0.31: icmp_seq=1 ttl=64 time=0.181 ms
64 bytes from 10.0.0.31: icmp_seq=2 ttl=64 time=0.136 ms
64 bytes from 10.0.0.31: icmp_seq=3 ttl=64 time=0.087 ms
64 bytes from 10.0.0.31: icmp_seq=4 ttl=64 time=0.142 ms

--- 10.0.0.31 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 2999ms
rtt min/avg/max/mdev = 0.087/0.136/0.181/0.035 ms

Iftop average: 550 Mb

Re: HA iSCSI Target with DRBD (2 node cluster how-to)

Posted: 2019/08/11 01:35:41
by hunter86_bg
Latency is OK, interface should be OK.

Most probaly it's a drbd sync limit.Check the link in my previous comment - it might help you speedup the situation.

Re: HA iSCSI Target with DRBD (2 node cluster how-to)

Posted: 2019/08/11 10:12:37
by ladam@ictuniverse.eu
Hi,

With this in the drbd0.res, I reach 1.5 GB in iftop.

disk { al-extents 3389;
no-disk-flushes ;
no-disk-barrier;
c-plan-ahead 50;
c-fill-target 24M;
c-min-rate 10M;
c-max-rate 1000M;
}


net { max-buffers 72000;
max-epoch-size 20000;
sndbuf-size 2M;
cram-hmac-alg sha1;
shared-secret "FooFunFactory";

}

Re: HA iSCSI Target with DRBD (2 node cluster how-to)

Posted: 2019/08/11 16:22:49
by hunter86_bg
ladam@ictuniverse.eu wrote:
2019/08/11 10:12:37
Hi,

With this in the drbd0.res, I reach 1.5 GB in iftop.

disk { al-extents 3389;
no-disk-flushes ;
no-disk-barrier;
c-plan-ahead 50;
c-fill-target 24M;
c-min-rate 10M;
c-max-rate 1000M;
}


net { max-buffers 72000;
max-epoch-size 20000;
sndbuf-size 2M;
cram-hmac-alg sha1;
shared-secret "FooFunFactory";

}
Great to hear that.
Just change the 'shared-secret' if you go over the internet... Well, it's always a good idea to chabge it - as its the default in the docs.

Re: HA iSCSI Target with DRBD (2 node cluster how-to)

Posted: 2019/08/11 16:58:20
by ladam@ictuniverse.eu
Hi,

So not so great. Rate was probably due to 2 drbd source at the same time.
Max is still 550 Mb.
each.
So I'm still looking because my goal is to sync 3 servers with 80 TB.
For the secret, yes but for now it's only on a private network.

Regards and thanks for your support

Re: HA iSCSI Target with DRBD (2 node cluster how-to)

Posted: 2019/08/11 18:51:06
by hunter86_bg
If you have 3 servers , you might consider GlusterFS with a 'replica 3' volume.

The setup is way easier.

Re: HA iSCSI Target with DRBD (2 node cluster how-to)

Posted: 2019/08/30 13:13:52
by nafiz
Thank you for this awesome and detailed implementation of DRBD with two node. It works !!!!! but unfortunately does not work with target. Works with tgtd. And also works in CentOS 7.6 smoothly.