Mailing List Archive

On Mon, Dec 14, 2020 at 03:43:18PM +0200, Pierre-Philipp Braun wrote:
> Hello
>
> imagine a three node cluster, with a resource `r1` configured across node1
> and 2. How to reach the very same resource but diskless on node3? I
> suppose I should look into NVME-over-TCP or iSCSI but it should be possible
> with DRBD itself, right? If so, would that be with both v8 and/or v9?

Use v9.

> With v9, I tried adding a `on node3` stanza for that resource, with `disk
> none`, but I get a warning saying that it has no "address" defined.

So did you define an address? :)

> As the
> goal is to reach the resource and not to share it, I don't know what to do.

Not entirely sure how you define "reach" and "share".

But yes, having it "disk none" on the diskless node is the way to go and
should work. For more I'd need to see the resource files, but in general
you are doing it right.

And while it is refreshing to see this 1 in a million person that
actually starts where one should start (at the bottom) and not on the
highest abstraction there is, I'd suggest you could still use LINSTOR,
create the resource on node 1 and 2 and on 3 diskless. And then look at
the resource files LINSTOR generated. That will not spoil you too much
:).

Regards, rck
_______________________________________________
Star us on GITHUB: https://github.com/LINBIT
drbd-user mailing list
drbd-user@lists.linbit.com
https://lists.linbit.com/mailman/listinfo/drbd-user

> So did you define an address? :)

Sorry, I was baffled with the way DRBD node communicate, and mistaken a
listening port with an SCSI target. It works! It's pretty cool)

I've tried with Protocol A and I could also mount a resource on node3
indeed. Hmm but what's the logic behind such a setup? As the primary
node here is diskless, the data is really unsecured and not written
anywhere, until it reaches the secondary nodes, right?

If I wanted to use that diskless trick, to make the resources available
across the whole cluster farm, I would probably have to use Protocol C
instead, in view not to loose any data in case of a node crash, correct?

BTW I noticed some misleading message on the diskless node while the
initial sync was on-going between the two other nodes. Not sure it's a
bug, but I mention it for the record.

r1 role:Secondary
disk:Diskless
lin1 role:Primary
peer-disk:UpToDate
lin2 role:Secondary
peer-disk:Inconsistent resync-suspended:peer

> create the resource on node 1 and 2 and on 3 diskless. And then look at
> the resource files LINSTOR generated. That will not spoil you too much
> :).

When creating a resource with --drbd-diskless, a totally useless and
empty resource gets created. So it's much worse than doing it manually.
I don't understand how to use that linstor option correctly, to begin
with.

Thank you
_______________________________________________
Star us on GITHUB: https://github.com/LINBIT
drbd-user mailing list
drbd-user@lists.linbit.com
https://lists.linbit.com/mailman/listinfo/drbd-user

Hi Pierre-Philipp,

Yes, Protocol C would be the most suitable here.

It means that the DRBD volume acknowledges the write after an actual write through on all diskful nodes has happened.

--
Best regards
Thomas Keppler

> Am 15.12.2020 um 11:03 schrieb Pierre-Philipp Braun <pbraun@nethence.com>:
>
> ?
>>
>> So did you define an address? :)
>
> Sorry, I was baffled with the way DRBD node communicate, and mistaken a listening port with an SCSI target. It works! It's pretty cool)
>
> I've tried with Protocol A and I could also mount a resource on node3 indeed. Hmm but what's the logic behind such a setup? As the primary node here is diskless, the data is really unsecured and not written anywhere, until it reaches the secondary nodes, right?
>
> If I wanted to use that diskless trick, to make the resources available across the whole cluster farm, I would probably have to use Protocol C instead, in view not to loose any data in case of a node crash, correct?
>
> BTW I noticed some misleading message on the diskless node while the initial sync was on-going between the two other nodes. Not sure it's a bug, but I mention it for the record.
>
> r1 role:Secondary
> disk:Diskless
> lin1 role:Primary
> peer-disk:UpToDate
> lin2 role:Secondary
> peer-disk:Inconsistent resync-suspended:peer
>
>> create the resource on node 1 and 2 and on 3 diskless. And then look at
>> the resource files LINSTOR generated. That will not spoil you too much
>> :).
>
> When creating a resource with --drbd-diskless, a totally useless and empty resource gets created. So it's much worse than doing it manually. I don't understand how to use that linstor option correctly, to begin with.
>
> Thank you
> _______________________________________________
> Star us on GITHUB: https://github.com/LINBIT
> drbd-user mailing list
> drbd-user@lists.linbit.com
> https://lists.linbit.com/mailman/listinfo/drbd-user
_______________________________________________
Star us on GITHUB: https://github.com/LINBIT
drbd-user mailing list
drbd-user@lists.linbit.com
https://lists.linbit.com/mailman/listinfo/drbd-user