[DRBD-user] Some question about linstor

[yang zhenyu]

Hi，Robert

Now the node status is ok, but integration zfs is still a problem, so I
still need your help.

Here are my steps:

[root at node-1 linstor-server]# linstor list-node
╭─────────────────────────────────────────╮
┊ Node ┊ NodeType ┊ IPs ┊ State ┊
╞┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄╡
┊ node-1 ┊ COMBINED ┊ 10.10.0.72 ┊ ok ┊
┊ node-4 ┊ SATELLITE ┊ 10.10.0.75 ┊ ok ┊
╰─────────────────────────────────────────╯

I hava a zfs pool named linstorpool.

[root at node-1 linstor-server]# zfs list
NAME USED AVAIL REFER MOUNTPOINT
linstorpool 384K 1.76T 96K /linstorpool

Then I create a storage-pool, I don't understand the args 'driver_device'
means what?
Is a zfs zvol name? or zvol device name? like zd0...

usage: linstor create-storage-pool [-h]
name node_name {lvm,lvmthin,zfs}
driver_device

Here I am free to use a name 'test1'.

linstor create-storage-pool zfs node-1 zfs test1
[root at node-1 linstor-server]# linstor list-storage-pool
╭─────────────────────────────────────────────────────────────────────╮
┊ StoragePool ┊ Node ┊ Driver ┊ DriverDevice ┊ SupportsSnapshots ┊
╞┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄╡
┊ zfs ┊ node-1 ┊ ZfsDriver ┊ test1 ┊ true ┊
╰─────────────────────────────────────────────────────────────────────╯
root at node-1 linstor-server]# linstor create-resource-definition rstdfn
[root at node-1 linstor-server]# linstor list-resource-definition
╭─────────────────────────────╮
┊ ResourceName ┊ Port ┊ State ┊
╞┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄╡
┊ rstdfn ┊ 7000 ┊ ok ┊
╰─────────────────────────────╯
[root at node-1 linstor-server]# linstor create-resource -s zfs -d rstdfn
node-1
[root at node-1 linstor-server]# linstor create-resource -s zfs -d rstdfn
node-4
[root at node-1 linstor-server]# linstor list-resource
╭─────────────────────────────────────────────────╮
┊ ResourceName ┊ Node ┊ Port ┊ Device ┊ State ┊
╞┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄╡
┊ rstdfn ┊ node-1 ┊ 7000 ┊ DISKLESS ┊ ok ┊
┊ rstdfn ┊ node-4 ┊ 7000 ┊ DISKLESS ┊ ok ┊
╰─────────────────────────────────────────────────╯

Why this Device is DISKLESS?

[root at node-1 linstor-server]# linstor create-volume-definition -d 2 rstdfn
30G
SUCCESS:
New volume definition with number '0' of resource definition 'rstdfn'
created.
WARNING:
Description:
Satellite reported an error during fullSync. This change will NOT be
delivered to satellte 'node-4' until the error is resolved. Reconnect the
satellite to the controller to remove this blockade.
Details:
Resource definition: rstdfn, Volume number: 0

[root at node-1 ~]# tailf /var/log/messages
Feb 23 17:38:52 node-1 Satellite: Adjusting resource rstdfn
Feb 23 17:38:52 node-1 kernel: drbd rstdfn: Starting worker thread (from
drbdsetup [4730])
Feb 23 17:38:52 node-1 kernel: drbd rstdfn node-4: Starting sender thread
(from drbdsetup [4735])
Feb 23 17:38:52 node-1 kernel: drbd rstdfn node-4: conn( StandAlone ->
Unconnected )
Feb 23 17:38:52 node-1 kernel: drbd rstdfn node-4: Starting receiver thread
(from drbd_w_rstdfn [4731])
Feb 23 17:38:52 node-1 kernel: drbd rstdfn node-4: conn( Unconnected ->
Connecting )

[root at node-1 linstor-server]# linstor list-volume
╭──────────────────────────────────────────────────────╮
┊ Node ┊ Resource ┊ VolumeNr ┊ MinorNr ┊ State ┊
╞┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄┄╡
┊ node-1 ┊ rstdfn ┊ 0 ┊ 1000 ┊ not present ┊
┊ node-4 ┊ rstdfn ┊ 0 ┊ 1000 ┊ unknown ┊
╰──────────────────────────────────────────────────────╯

[root at node-1 linstor-server]# drbdadm status
rstdfn role:Secondary
disk:Diskless
node-4 connection:Connecting

[root at node-4 opt]# drbdadm status
# No currently configured DRBD found.

[root at node-4 opt]# cat /etc/drbd.d/rstdfn.res
cat: /etc/drbd.d/rstdfn.res: No such file or directory

I do not know where is the problem. Is it network communication problems?
But I `telnet node-4 3366` is no problem.

2018-02-22 19:02 GMT+08:00 Robert Altnoeder <robert.altnoeder at linbit.com>:

> Hello,
>
> On 02/22/2018 11:35 AM, yang zhenyu wrote:
> > [...]
> >
> > The following is my understanding:
> >
> > 1. The LINSTOR Controller node corresponds to the drbdmanage pure
> > controller, there is no storage.
> > 2. The LINSTOR Combined node corresponds to the drbdmanage controller
> > with storage.
> > 3. The LINSTOR Satellites node  corresponds to the drbdmanage
> > satellites with storage.
>
> LINSTOR comprises two modules, the controller and the satellite.
>
> Only the controller manages the configuration, e.g., all resource
> definitions, storage pools, network interfaces of nodes, etc. are
> defined on the controller.
>
> Only the satellites deal with actual storage, with DRBD, etc., and the
> satellites obtain information about the configuration of their resources
> from the active controller (the controller contacts all registered
> satellites).
>
> Only one controller is active at a time, but multiple installations of
> the controller may be provided so that if the active controller node
> crashes, the controller can be started on another node. This kind of
> failover is managed by a cluster manager such as Pacemaker.
>
> A combined node is a node that can run the controller module and the
> satellite module. This is required if one of the storage nodes is also
> running the controller module. In this case, the controller module and
> the satellite module will be running on that node at the same time, and
> the controller will connect to the satellite on the local node the same
> way it connects to remote satellites.
>
> The controller can also be installed on platforms that do not support
> the kind of storage that the satellites support, or that do not support
> DRBD (e.g., you could run the controller on an AIX or HP/UX box and have
> it control satellites that run on Linux, and in the future also Windows
> with DRBD for Windows).
>
> > 4. The LINSTOR Auxiliary node corresponds to the drbdmanage client node?
>
> Auxiliary nodes are more or less disabled nodes, they do not run the
> LINSTOR software. They are needed for future features, as of now, they
> are not very useful for anything.
>
> > 5. Only the control node to access the database?
>
> Only one controller is connected to the database at any time.
>
> > Currently linstor only use the derby database? So I need a shared
> storage for multiple
> > control nodes.
>
> Technically, the controller is able to connect to a
> centralized/networked database (we run one of our development systems
> with such a configuration), but normally it runs an Apache Derby
> embedded database in the controller module's process. The database layer
> is exchangable, but Derby is the only driver that is implemented as of
> today.
>
> > But I do not understand what is the ues of cluster manage software at
> here? For example pacemaker.
>
> It can be used for automatic failover to another controller if the
> active controller node fails.
>
> best regards,
> --
> Robert Altnoeder
> +43 1 817 82 92 0
> robert.altnoeder at linbit.com
>
> LINBIT | Keeping The Digital World Running
> DRBD - Corosync - Pacemaker
> f /  t /  in /  g+
>
> DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20180223/f83a71f7/attachment-0001.htm>