You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@cloudstack.apache.org by Bryan Tiang <br...@hotmail.com> on 2023/10/18 10:11:12 UTC

Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Hi Guys,

We are doing some evaluation with Cloudstack + Linbit SDS.

Has anyone had any experience using these with a Converged or Hyperconverged setup?

My understanding is that Converged is the best for HA Because:

• If any storage node goes down, there is zero downtime. (3 Time Replica)
• If any compute node goes down, it will be restarted in another node as part of HA feature.

But what about Hyperconverged setup? Can we also set zero downtime with storage and fast VM recovery?

Regards,
Bryan

Re: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Posted by Dilip Singh <di...@datahub.com.np>.
unsubscribe.

*From: *Bryan Tiang [mailto:bryantiang15@hotmail.com]
*Sent: *Wednesday, October 18, 2023 03:56 PM +5.75
*To: *users@cloudstack.apache.org
*Subject: *Comparing Hyperconverged + Converged Setup with Cloudstack + 
Linbit

Hi Guys,

We are doing some evaluation with Cloudstack + Linbit SDS.

Has anyone had any experience using these with a Converged or Hyperconverged setup?

My understanding is that Converged is the best for HA Because:

• If any storage node goes down, there is zero downtime. (3 Time Replica)
• If any compute node goes down, it will be restarted in another node as part of HA feature.

But what about Hyperconverged setup? Can we also set zero downtime with storage and fast VM recovery?

Regards,
Bryan

AW: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Posted by me...@swen.io.
Interesting, did you be able to reproduce this failing? Maybe we should talked directly, because this is not really part of Cloudstack.

-----Ursprüngliche Nachricht-----
Von: Pratik Chandrakar <ch...@gmail.com> 
Gesendet: Freitag, 3. November 2023 12:11
An: users@cloudstack.apache.org
Betreff: Re: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Hi Swen,
We also used drbd-reactor, but it failed twice in our testing. So we stopped it for further testing.

On Fri, Nov 3, 2023 at 3:45 PM <me...@swen.io> wrote:

> Hi Pratik,
>
> How do you want to achieve HA for the controller? We are using 
> drbd-reactor and it worked everytime and never failed in our tests.
>
> Regards,
> Swen
>
> -----Ursprüngliche Nachricht-----
> Von: Pratik Chandrakar <ch...@gmail.com>
> Gesendet: Freitag, 3. November 2023 06:25
> An: users@cloudstack.apache.org
> Betreff: Re: Comparing Hyperconverged + Converged Setup with 
> Cloudstack + Linbit
>
> Hi Swen,
> After enabling HA for Linstor controller, it failed to start from the 
> secondary node twice and also failed to start after rebooting the 
> existing controller.
>
> On Thu, Oct 19, 2023 at 12:48 PM <me...@swen.io> wrote:
>
> > Hey Pratik,
> >
> > can you elaborate more on this stability problems? We are doing also 
> > a CS
> > + Linstor PoC at the moment and we did a lot of stress testing it 
> > + without
> > any problems on linstor side. I am curious if we did miss some tests.
> > We are using a place count of 2 in a 3 node cluster.
> >
> > Regards,
> > Swen
> >
> > -----Ursprüngliche Nachricht-----
> > Von: Pratik Chandrakar <ch...@gmail.com>
> > Gesendet: Donnerstag, 19. Oktober 2023 07:15
> > An: users@cloudstack.apache.org
> > Betreff: Re: Comparing Hyperconverged + Converged Setup with 
> > Cloudstack + Linbit
> >
> > Hi Bryan,
> >
> > We did a small PoC with Cloudstack + Linbit SDS(3 Time replica) in a 
> > hyperconverged setup. There was no issue with HA, the VMs 
> > successfully restarted from different nodes. However, we did face 
> > stability problems with Linbit HA, which prevented us from 
> > provisioning new storage or virtual machines.
> >
> > On Wed, Oct 18, 2023 at 3:42 PM Bryan Tiang 
> > <br...@hotmail.com>
> > wrote:
> >
> > > Hi Guys,
> > >
> > > We are doing some evaluation with Cloudstack + Linbit SDS.
> > >
> > > Has anyone had any experience using these with a Converged or 
> > > Hyperconverged setup?
> > >
> > > My understanding is that Converged is the best for HA Because:
> > >
> > > • If any storage node goes down, there is zero downtime. (3 Time
> > > Replica) • If any compute node goes down, it will be restarted in 
> > > another node as part of HA feature.
> > >
> > > But what about Hyperconverged setup? Can we also set zero downtime 
> > > with storage and fast VM recovery?
> > >
> > > Regards,
> > > Bryan
> > >
> >
> >
> > --
> > *Regards,*
> > *Pratik Chandrakar*
> >
> >
> >
>
> --
> *Regards,*
> *Pratik Chandrakar*
>
>
>

--
*Regards,*
*Pratik Chandrakar*



Re: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Posted by Pratik Chandrakar <ch...@gmail.com>.
Hi Swen,
We also used drbd-reactor, but it failed twice in our testing. So we
stopped it for further testing.

On Fri, Nov 3, 2023 at 3:45 PM <me...@swen.io> wrote:

> Hi Pratik,
>
> How do you want to achieve HA for the controller? We are using
> drbd-reactor and it worked everytime and never failed in our tests.
>
> Regards,
> Swen
>
> -----Ursprüngliche Nachricht-----
> Von: Pratik Chandrakar <ch...@gmail.com>
> Gesendet: Freitag, 3. November 2023 06:25
> An: users@cloudstack.apache.org
> Betreff: Re: Comparing Hyperconverged + Converged Setup with Cloudstack +
> Linbit
>
> Hi Swen,
> After enabling HA for Linstor controller, it failed to start from the
> secondary node twice and also failed to start after rebooting the existing
> controller.
>
> On Thu, Oct 19, 2023 at 12:48 PM <me...@swen.io> wrote:
>
> > Hey Pratik,
> >
> > can you elaborate more on this stability problems? We are doing also a
> > CS
> > + Linstor PoC at the moment and we did a lot of stress testing it
> > + without
> > any problems on linstor side. I am curious if we did miss some tests.
> > We are using a place count of 2 in a 3 node cluster.
> >
> > Regards,
> > Swen
> >
> > -----Ursprüngliche Nachricht-----
> > Von: Pratik Chandrakar <ch...@gmail.com>
> > Gesendet: Donnerstag, 19. Oktober 2023 07:15
> > An: users@cloudstack.apache.org
> > Betreff: Re: Comparing Hyperconverged + Converged Setup with
> > Cloudstack + Linbit
> >
> > Hi Bryan,
> >
> > We did a small PoC with Cloudstack + Linbit SDS(3 Time replica) in a
> > hyperconverged setup. There was no issue with HA, the VMs successfully
> > restarted from different nodes. However, we did face stability
> > problems with Linbit HA, which prevented us from provisioning new
> > storage or virtual machines.
> >
> > On Wed, Oct 18, 2023 at 3:42 PM Bryan Tiang <br...@hotmail.com>
> > wrote:
> >
> > > Hi Guys,
> > >
> > > We are doing some evaluation with Cloudstack + Linbit SDS.
> > >
> > > Has anyone had any experience using these with a Converged or
> > > Hyperconverged setup?
> > >
> > > My understanding is that Converged is the best for HA Because:
> > >
> > > • If any storage node goes down, there is zero downtime. (3 Time
> > > Replica) • If any compute node goes down, it will be restarted in
> > > another node as part of HA feature.
> > >
> > > But what about Hyperconverged setup? Can we also set zero downtime
> > > with storage and fast VM recovery?
> > >
> > > Regards,
> > > Bryan
> > >
> >
> >
> > --
> > *Regards,*
> > *Pratik Chandrakar*
> >
> >
> >
>
> --
> *Regards,*
> *Pratik Chandrakar*
>
>
>

-- 
*Regards,*
*Pratik Chandrakar*

AW: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Posted by me...@swen.io.
Hi Pratik,

How do you want to achieve HA for the controller? We are using drbd-reactor and it worked everytime and never failed in our tests.

Regards,
Swen

-----Ursprüngliche Nachricht-----
Von: Pratik Chandrakar <ch...@gmail.com> 
Gesendet: Freitag, 3. November 2023 06:25
An: users@cloudstack.apache.org
Betreff: Re: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Hi Swen,
After enabling HA for Linstor controller, it failed to start from the secondary node twice and also failed to start after rebooting the existing controller.

On Thu, Oct 19, 2023 at 12:48 PM <me...@swen.io> wrote:

> Hey Pratik,
>
> can you elaborate more on this stability problems? We are doing also a 
> CS
> + Linstor PoC at the moment and we did a lot of stress testing it 
> + without
> any problems on linstor side. I am curious if we did miss some tests.
> We are using a place count of 2 in a 3 node cluster.
>
> Regards,
> Swen
>
> -----Ursprüngliche Nachricht-----
> Von: Pratik Chandrakar <ch...@gmail.com>
> Gesendet: Donnerstag, 19. Oktober 2023 07:15
> An: users@cloudstack.apache.org
> Betreff: Re: Comparing Hyperconverged + Converged Setup with 
> Cloudstack + Linbit
>
> Hi Bryan,
>
> We did a small PoC with Cloudstack + Linbit SDS(3 Time replica) in a 
> hyperconverged setup. There was no issue with HA, the VMs successfully 
> restarted from different nodes. However, we did face stability 
> problems with Linbit HA, which prevented us from provisioning new 
> storage or virtual machines.
>
> On Wed, Oct 18, 2023 at 3:42 PM Bryan Tiang <br...@hotmail.com>
> wrote:
>
> > Hi Guys,
> >
> > We are doing some evaluation with Cloudstack + Linbit SDS.
> >
> > Has anyone had any experience using these with a Converged or 
> > Hyperconverged setup?
> >
> > My understanding is that Converged is the best for HA Because:
> >
> > • If any storage node goes down, there is zero downtime. (3 Time
> > Replica) • If any compute node goes down, it will be restarted in 
> > another node as part of HA feature.
> >
> > But what about Hyperconverged setup? Can we also set zero downtime 
> > with storage and fast VM recovery?
> >
> > Regards,
> > Bryan
> >
>
>
> --
> *Regards,*
> *Pratik Chandrakar*
>
>
>

--
*Regards,*
*Pratik Chandrakar*



Re: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Posted by Pratik Chandrakar <ch...@gmail.com>.
Hi Swen,
After enabling HA for Linstor controller, it failed to start from the
secondary node twice and also failed to start after rebooting the existing
controller.

On Thu, Oct 19, 2023 at 12:48 PM <me...@swen.io> wrote:

> Hey Pratik,
>
> can you elaborate more on this stability problems? We are doing also a CS
> + Linstor PoC at the moment and we did a lot of stress testing it without
> any problems on linstor side. I am curious if we did miss some tests.
> We are using a place count of 2 in a 3 node cluster.
>
> Regards,
> Swen
>
> -----Ursprüngliche Nachricht-----
> Von: Pratik Chandrakar <ch...@gmail.com>
> Gesendet: Donnerstag, 19. Oktober 2023 07:15
> An: users@cloudstack.apache.org
> Betreff: Re: Comparing Hyperconverged + Converged Setup with Cloudstack +
> Linbit
>
> Hi Bryan,
>
> We did a small PoC with Cloudstack + Linbit SDS(3 Time replica) in a
> hyperconverged setup. There was no issue with HA, the VMs successfully
> restarted from different nodes. However, we did face stability problems
> with Linbit HA, which prevented us from provisioning new storage or virtual
> machines.
>
> On Wed, Oct 18, 2023 at 3:42 PM Bryan Tiang <br...@hotmail.com>
> wrote:
>
> > Hi Guys,
> >
> > We are doing some evaluation with Cloudstack + Linbit SDS.
> >
> > Has anyone had any experience using these with a Converged or
> > Hyperconverged setup?
> >
> > My understanding is that Converged is the best for HA Because:
> >
> > • If any storage node goes down, there is zero downtime. (3 Time
> > Replica) • If any compute node goes down, it will be restarted in
> > another node as part of HA feature.
> >
> > But what about Hyperconverged setup? Can we also set zero downtime
> > with storage and fast VM recovery?
> >
> > Regards,
> > Bryan
> >
>
>
> --
> *Regards,*
> *Pratik Chandrakar*
>
>
>

-- 
*Regards,*
*Pratik Chandrakar*

Re: AW: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Posted by Pratik Chandrakar <ch...@gmail.com>.
Hi Bryan,
Sorry for the late reply. Yes we had zero downtime but no provisioning of
new machines got affected.

On Thu, Oct 19, 2023 at 1:48 PM Bryan Tiang <br...@hotmail.com>
wrote:

> Hi Pratik,
>
> Thanks for the response.
>
> Just to confirm, your storage volumes had zero downtime, right?
>
> Regards,
> Bryan
> On 19 Oct 2023 at 3:18 PM +0800, me@swen.io, wrote:
> > Hey Pratik,
> >
> > can you elaborate more on this stability problems? We are doing also a
> CS + Linstor PoC at the moment and we did a lot of stress testing it
> without any problems on linstor side. I am curious if we did miss some
> tests.
> > We are using a place count of 2 in a 3 node cluster.
> >
> > Regards,
> > Swen
> >
> > -----Ursprüngliche Nachricht-----
> > Von: Pratik Chandrakar <ch...@gmail.com>
> > Gesendet: Donnerstag, 19. Oktober 2023 07:15
> > An: users@cloudstack.apache.org
> > Betreff: Re: Comparing Hyperconverged + Converged Setup with Cloudstack
> + Linbit
> >
> > Hi Bryan,
> >
> > We did a small PoC with Cloudstack + Linbit SDS(3 Time replica) in a
> hyperconverged setup. There was no issue with HA, the VMs successfully
> restarted from different nodes. However, we did face stability problems
> with Linbit HA, which prevented us from provisioning new storage or virtual
> machines.
> >
> > On Wed, Oct 18, 2023 at 3:42 PM Bryan Tiang <br...@hotmail.com>
> > wrote:
> >
> > > Hi Guys,
> > >
> > > We are doing some evaluation with Cloudstack + Linbit SDS.
> > >
> > > Has anyone had any experience using these with a Converged or
> > > Hyperconverged setup?
> > >
> > > My understanding is that Converged is the best for HA Because:
> > >
> > > • If any storage node goes down, there is zero downtime. (3 Time
> > > Replica) • If any compute node goes down, it will be restarted in
> > > another node as part of HA feature.
> > >
> > > But what about Hyperconverged setup? Can we also set zero downtime
> > > with storage and fast VM recovery?
> > >
> > > Regards,
> > > Bryan
> > >
> >
> >
> > --
> > *Regards,*
> > *Pratik Chandrakar*
> >
> >
>


-- 
*Regards,*
*Pratik Chandrakar*

Re: AW: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Posted by Bryan Tiang <br...@hotmail.com>.
Hi Pratik,

Thanks for the response.

Just to confirm, your storage volumes had zero downtime, right?

Regards,
Bryan
On 19 Oct 2023 at 3:18 PM +0800, me@swen.io, wrote:
> Hey Pratik,
>
> can you elaborate more on this stability problems? We are doing also a CS + Linstor PoC at the moment and we did a lot of stress testing it without any problems on linstor side. I am curious if we did miss some tests.
> We are using a place count of 2 in a 3 node cluster.
>
> Regards,
> Swen
>
> -----Ursprüngliche Nachricht-----
> Von: Pratik Chandrakar <ch...@gmail.com>
> Gesendet: Donnerstag, 19. Oktober 2023 07:15
> An: users@cloudstack.apache.org
> Betreff: Re: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit
>
> Hi Bryan,
>
> We did a small PoC with Cloudstack + Linbit SDS(3 Time replica) in a hyperconverged setup. There was no issue with HA, the VMs successfully restarted from different nodes. However, we did face stability problems with Linbit HA, which prevented us from provisioning new storage or virtual machines.
>
> On Wed, Oct 18, 2023 at 3:42 PM Bryan Tiang <br...@hotmail.com>
> wrote:
>
> > Hi Guys,
> >
> > We are doing some evaluation with Cloudstack + Linbit SDS.
> >
> > Has anyone had any experience using these with a Converged or
> > Hyperconverged setup?
> >
> > My understanding is that Converged is the best for HA Because:
> >
> > • If any storage node goes down, there is zero downtime. (3 Time
> > Replica) • If any compute node goes down, it will be restarted in
> > another node as part of HA feature.
> >
> > But what about Hyperconverged setup? Can we also set zero downtime
> > with storage and fast VM recovery?
> >
> > Regards,
> > Bryan
> >
>
>
> --
> *Regards,*
> *Pratik Chandrakar*
>
>

AW: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Posted by me...@swen.io.
Hey Pratik,

can you elaborate more on this stability problems? We are doing also a CS + Linstor PoC at the moment and we did a lot of stress testing it without any problems on linstor side. I am curious if we did miss some tests.
We are using a place count of 2 in a 3 node cluster.

Regards,
Swen

-----Ursprüngliche Nachricht-----
Von: Pratik Chandrakar <ch...@gmail.com> 
Gesendet: Donnerstag, 19. Oktober 2023 07:15
An: users@cloudstack.apache.org
Betreff: Re: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Hi Bryan,

We did a small PoC with Cloudstack + Linbit SDS(3 Time replica) in a hyperconverged setup. There was no issue with HA, the VMs successfully restarted from different nodes. However, we did face stability problems with Linbit HA, which prevented us from provisioning new storage or virtual machines.

On Wed, Oct 18, 2023 at 3:42 PM Bryan Tiang <br...@hotmail.com>
wrote:

> Hi Guys,
>
> We are doing some evaluation with Cloudstack + Linbit SDS.
>
> Has anyone had any experience using these with a Converged or 
> Hyperconverged setup?
>
> My understanding is that Converged is the best for HA Because:
>
> • If any storage node goes down, there is zero downtime. (3 Time 
> Replica) • If any compute node goes down, it will be restarted in 
> another node as part of HA feature.
>
> But what about Hyperconverged setup? Can we also set zero downtime 
> with storage and fast VM recovery?
>
> Regards,
> Bryan
>


--
*Regards,*
*Pratik Chandrakar*



Re: Comparing Hyperconverged + Converged Setup with Cloudstack + Linbit

Posted by Pratik Chandrakar <ch...@gmail.com>.
Hi Bryan,

We did a small PoC with Cloudstack + Linbit SDS(3 Time replica) in a
hyperconverged setup. There was no issue with HA, the VMs successfully
restarted from different nodes. However, we did face stability problems
with Linbit HA, which prevented us from provisioning new storage or virtual
machines.

On Wed, Oct 18, 2023 at 3:42 PM Bryan Tiang <br...@hotmail.com>
wrote:

> Hi Guys,
>
> We are doing some evaluation with Cloudstack + Linbit SDS.
>
> Has anyone had any experience using these with a Converged or
> Hyperconverged setup?
>
> My understanding is that Converged is the best for HA Because:
>
> • If any storage node goes down, there is zero downtime. (3 Time Replica)
> • If any compute node goes down, it will be restarted in another node as
> part of HA feature.
>
> But what about Hyperconverged setup? Can we also set zero downtime with
> storage and fast VM recovery?
>
> Regards,
> Bryan
>


-- 
*Regards,*
*Pratik Chandrakar*