[DRBD-user] (newbie question) DRBD 9, pure controller without data storage

Riches Jr, Robert M robert.m.riches.jr at intel.com
Tue Dec 5 19:28:05 CET 2017

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


(Follow-up question below.  Apologies for the inconsistent quoting style.  I don't know
how to force Outlook to do more consistent quoting.)

-----Original Message-----
From: drbd-user-bounces at lists.linbit.com [mailto:drbd-user-bounces at lists.linbit.com] On Behalf Of Roland Kammerer
Sent: Tuesday, December 05, 2017 1:17 AM
To: drbd-user at lists.linbit.com
Subject: Re: [DRBD-user] (newbie question) DRBD 9, pure controller without data storage

On Mon, Dec 04, 2017 at 11:10:16PM +0000, Riches Jr, Robert M wrote:
> (This is a newbie question.)
> 
> Having got non-LVM version 9 drbdadm working with two storage nodes 
> and a few no-local-storage nodes, I'm attempting to graduate to using 
> drbdmanage and LVM.  My goal is to have two storage nodes, one pure 
> controller node without data storage, and a few no-local-storage 
> compute nodes.  However, I don't understand the green box in Figure 5 
> in the User's Guide, labeled as a 'Control node', 'controlvolume is a 
> local LV', 'pure Controller', and 'nodes without data storage'.  This is the figure:
> 
> https://docs.linbit.com/ug-src/users-guide-9.0/images/drbdmanage-venn.
> png
> 
> The only LVM volume group I see mentioned in chapter 5 of the DRBD 9 
> UG is the one it suggests to name 'drbdpool'.  However, if the pure 
> controller has no data storage, how can the controlvolume be a local 
> LV.  Can someone please enlighten a newbie on what seem to be contradictory descriptions?

There are two types of storage drbdmange can use:
- the control volume (which is the cluster wide database, 2x4MB, always
  on thick LVM)
- the storage for your actual DRBD resources (LVM/ZFS thin/thick).

That makes 4 combinations:
- The node has a copy of the control volume, that makes it a "control
  node". It can then provide storage for DRBD resources or not (control
  node or pure controller). A use case for a pure controller might be a
  "quorum node" that just has a copy of the cluster DB. Let's say you
  have 3 nodes, one of them is a pure controller, one of the other nodes
  fails then the cluster can continue as it still has 2 nodes with a
  copy of the control volume.
- The node does not have a copy of the cluster DB and gets its cluster
  DB from another node via TCP. These are satellites. They can have
  local storage for DRBD resources or not.

Regards, rck
-----End Of Original Message-----

Thank you for that explanation/clarification.  I think I understand now.  Let
me make sure my understanding is correct, especially concerning the size and
contents of the VGs:

- A "quorum node" will have a VG named 'drbdpool' that needs only enough space
for the control volume.  It will be added by doing
"drbdmanage add-node --no-storage <nodeName> [IPaddress]"

- Each other "control node" will have a VG named 'drbdpool' with enough space
for the control volume _and_  the DRBD resources.  It will be added by doing
"drbdmanage add-node <nodeName> [IPaddress]" (no options).

- A "pure client" node will have _no_ VG.  It will be added by doing
"drbdmanage add-node --satellite --no-storage <nodeName> [IPaddress]".

- If I determine that an external node is one of the droids I'm looking for,
it will have _no_ VG.  It will be added by doing
"drbdmanage add-node --external <nodeName> [IPaddress]".

Is that correct?

Thanks,

Robert Riches




More information about the drbd-user mailing list