[DRBD-user] Best practice: drbd+lvm+gfs2+dm-crypt on dual primary

Digimer lists at alteeve.ca
Mon Feb 2 19:21:35 CET 2015

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

Hi Patrick,

   Please keep replies on the list, it helps keep the conversation in 
the archives which can help people in the future. :)

   If you get a "reboot without logging", please tell us exactly what 
triggered it and what your configuration is. Either it's a 
misconfiguration or a bug, either of which are good to discover and 
understand properly.

   To answer your questions;

   I use the switches to help with debugging. With back-to-back, if a 
link drops, both go down. It makes it harder to diagnose where the 
problem is. Also, with managed switches, you can get tools to test and 
diagnose links as well.

   Gluster is really a cloud solution, not really an HA solution, so I 
don't think it's relevant here (though others may disagree with me).

   Note as well that I only use active/passive (mode=1) bonds because, 
in my testing, that is the only method that can reliably survive both 
failure and recovery and has the fastest fault detection and recovery 



On 02/02/15 01:13 PM, Patrick Prilisauer wrote:
> Thanks!!,
> I'll give my Centos 7 setup one more chance, If that dumb thing, reboots
> again with no logged event I'll performe your tutorial.
> Why do you use a switch in a two node setup? I've used in my last setup
> Fibre channel single mode with an peer to peer connection, in fact I'm
> not using a system like Gluster,... Im only interessted in a robust
> storage backbone and the second connection for the Servers.
> is there another reason, than using the switch for sniffing?
> Anyway thanks a lot,..
> 2015-02-02 18:28 GMT+01:00 Digimer <lists at alteeve.ca
> <mailto:lists at alteeve.ca>>:
>     I show how to use DRBD + clvmd + gfs2 in this tutorial:
>     https://alteeve.ca/w/AN!__Cluster_Tutorial_2
>     <https://alteeve.ca/w/AN!Cluster_Tutorial_2>
>     I have clusters built following that tutorial all over north america
>     (and I know of users around the world). Since 2010, not a single
>     system was lost. The trick is that a lot of people don't understand
>     the critical role of fencing, disable it or install it broken, then
>     things break and they think it's the software. It's like blaming a
>     car for being unsafe when you don't wear your seatbelt. :)
>     digimer

Papers and Projects: https://alteeve.ca/w/
What if the cure for cancer is trapped in the mind of a person without 
access to education?

More information about the drbd-user mailing list