[PATCH docs] fix #5273: pvecm: qdevice: note to not run qnetd on the cluster itself

[Michal Fox <me@dualfroz.com>]

The section about the external vote support already says that the vote
daemon runs on an independent server, but does not spell out that this
rules out running it on the cluster itself, for example on one of the
nodes or in a guest, which might look like an easy option for a cluster
with only two nodes.

Add a note that says so explicitly, and why: if the node running the
daemon fails, the cluster loses two votes at once, and a guest running
it cannot be started without quorum, as starting a guest requires a
quorate cluster.

Signed-off-by: Michal Fox <me@dualfroz.com>
---

Notes:
    that guests cannot be started without quorum is based on the quorum
    check in the start API calls of qemu-server and pve-container. rendered
    the chapter with asciidoc, without new warnings.

 pvecm.adoc | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/pvecm.adoc b/pvecm.adoc
index 8309647..ac4e92f 100644
--- a/pvecm.adoc
+++ b/pvecm.adoc
@@ -1132,6 +1132,13 @@ For this to work, there are two services involved:
 As a result, you can achieve higher availability, even in smaller setups (for
 example 2+1 nodes).
 
+NOTE: The server running the external vote daemon must not be part of the
+cluster in any form. Do not run the daemon on a node of the same cluster, nor
+in a virtual machine or container on it. Otherwise, if the node running the
+daemon fails, the cluster loses the vote of this node and the one of the
+QDevice at the same time. Also, a guest running the daemon cannot be started
+while the cluster is not quorate.
+
 QDevice Technical Overview
 ~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-- 
2.43.0