samba - Jun 2024 - PVE HA for DCs

Dear Samba users.

Lately I am running a 3 node Proxmox cluster with??some VMs and CTs including
Samba DCs. I have to say I quite impressed with the performance and features.

Some of my CTs have HA turned on, basically, if a node in the cluster goes down
or gets isolated, PVE fences the node and moves the HA-configured CTs and VMs to
other hosts. I am using CEPH shared storage. As I understand this, basically the
VM or CT is restarted in the new node, with whatever content in on the shared
disk.

Works fine for example with DNS servers, radius serves, other auxiliary
services, etc.

However I am concerned this may break the AD if a DC is restarted like this, in
terms of databases syncing between DCs.

This question would be the same: If a DC crashed, can it be restarted in the AD
without any issues ? I think is a similar situation. I would say no, better keek
isolated, demote and re-join.

What are your thoughts ?

LP

Maybe I?ll rephrase the question, it should be equivalent to:

If a DC crashes and needs restarting, is it safe to do so, or because
replication has been interrupted and may be inconsistent, shall I isolate,
demote (off line demote) , and re-join ?

Thanks,

LP
On Jun 3, 2024 at 18:10 +0100, Luis Peromarta <lperoma at icloud.com>,
wrote:
> Dear Samba users.
>
> Lately I am running a 3 node Proxmox cluster with??some VMs and CTs
including Samba DCs. I have to say I quite impressed with the performance and
features.
>
> Some of my CTs have HA turned on, basically, if a node in the cluster goes
down or gets isolated, PVE fences the node and moves the HA-configured CTs and
VMs to other hosts. I am using CEPH shared storage. As I understand this,
basically the VM or CT is restarted in the new node, with whatever content in on
the shared disk.
>
> Works fine for example with DNS servers, radius serves, other auxiliary
services, etc.
>
> However I am concerned this may break the AD if a DC is restarted like
this, in terms of databases syncing between DCs.
>
> This question would be the same: If a DC crashed, can it be restarted in
the AD without any issues ? I think is a similar situation. I would say no,
better keek isolated, demote and re-join.
>
> What are your thoughts ?
>
> LP