XOSTOR hyperconvergence preview

ronan-a

@gb-123 said in XOSTOR hyperconvergence preview:

@ronan-a

running linstor command on both hosts gives the following error :

Traceback (most recent call last):
File "/usr/bin/linstor", line 21, in <module>
import linstor_client_main
ImportError: No module named linstor_client_main

Regarding this error, can you say me what's your XCP-ng version? And what's the output of: yum list installed | grep -i linstor? Do you have a XCP-ng 8.3? And did you run the beta installation script on it?

Also, your SSL/VLAN issue is not caused by LINSTOR, @BenjiReis do you have any idea?

gb.123

@ronan-a said in XOSTOR hyperconvergence preview:

@gb-123 said in XOSTOR hyperconvergence preview:

@ronan-a

running linstor command on both hosts gives the following error :

Traceback (most recent call last):
File "/usr/bin/linstor", line 21, in <module>
import linstor_client_main
ImportError: No module named linstor_client_main

Regarding this error, can you say me what's your XCP-ng version? And what's the output of: yum list installed | grep -i linstor?

Output of yum list installed | grep -i linstor :

drbd.x86_64                     9.22.0-1.el7               @xcp-ng-linstor      
drbd-bash-completion.x86_64     9.22.0-1.el7               @xcp-ng-linstor      
drbd-pacemaker.x86_64           9.22.0-1.el7               @xcp-ng-linstor      
drbd-reactor.x86_64             1.0.0-1                    @xcp-ng-linstor      
drbd-udev.x86_64                9.22.0-1.el7               @xcp-ng-linstor      
drbd-utils.x86_64               9.22.0-1.el7               @xcp-ng-linstor      
drbd-xen.x86_64                 9.22.0-1.el7               @xcp-ng-linstor      
kmod-drbd.x86_64                9.2.2+ptf.1_4.19.0+1-1     @xcp-ng-linstor      
linstor-client.noarch           1.18.0-1                   @xcp-ng-linstor      
linstor-common.noarch           1.21.1-1.el7               @xcp-ng-linstor      
linstor-controller.noarch       1.21.1-1.el7               @xcp-ng-linstor      
linstor-satellite.noarch        1.21.1-1.el7               @xcp-ng-linstor      
python-linstor.noarch           1.18.0-1                   @xcp-ng-linstor      
xcp-ng-linstor.noarch           1.2-1.xcpng8.3             @xcp-ng-linstor      
xcp-ng-release-linstor.noarch   1.4-1.xcpng8.3             @xcp-ng-base

Do you have a XCP-ng 8.3? And did you run the beta installation script on it?

I installed XCP-ng 8.3 beta from ISO and then applied all the patches that were available. To install linstor, I followed the 1st Post of this topic.

Also, your SSL/VLAN issue is not caused by LINSTOR, @BenjiReis do you have any idea?

Yeah, even I thought so, but nothing changed for me before I installed Linstor.

There is also a very peculiar thing I am noticing:
Backup from Thin (Ext4) SR to another Thin (Ext4) SR has a transfer of 120KiBs over network(WAN). I thought this is maybe due to a 'slow' network, but however, I started another backup at the same time while the previous one was running slowly, this time from Thick (LVM) SR to Thin SR and the speed I got was around 16 MiB/s (which seems to be OK). I did write a post regarding this but deleted it since I need to dig deeper into this before I report this as an issue to you.
I will be conducting tests for a few more days just to be sure. (I am talking about Continuous Backup btw.)

Update :
After a week of continuous testing, I am getting mixed results on the backup speed as mentioned in the above paragraph. I can now say that the issue may not be related to Linstor.

Maelstrom96

Is there a procedure on how we can update our current 8.2 XCP-ng cluster to 8.3? My undertanding is that if I update the host using the ISO, it will effectively wipe all changes that were made to DOM0, including the linstor/sm-linstor packages.

BenjiReis

@gb-123 are you using XOA or XO from sources?
If from sources the issue might come from having a different openssl when creating the sdn-controller certificates.
You can either try with an XOA or generate manually your certificates.

gb.123

@BenjiReis said in XOSTOR hyperconvergence preview:

@gb-123 are you using XOA or XO from sources?
If from sources the issue might come from having a different openssl when creating the sdn-controller certificates.
You can either try with an XOA or generate manually your certificates.

Using XO from sources. I just turned on "Over-ride certificates" and reinstalled the whole XO virtual machine. Seems to work fine now.

My only problem was that why it suddenly stopped working when I installed Linstore as installing Linstor should not impact upon this. So thats why I reported this on this thread.

BenjiReis

@gb-123 I see thanks

Just bad timing imho because Linstor doesn't touch this part of the host and the openssl issue is more probably coming from the env when you run your XO.

anyway, glad it's working now!

gb.123

@ronan-a

Any idea why I am getting the error while running the linstor command ?

Thanks

ronan-a

@gb-123 What command and error exactly?

Maelstrom96

@Maelstrom96 said in XOSTOR hyperconvergence preview:

Is there a procedure on how we can update our current 8.2 XCP-ng cluster to 8.3? My undertanding is that if I update the host using the ISO, it will effectively wipe all changes that were made to DOM0, including the linstor/sm-linstor packages.

Any input on this @ronan-a?

olivierlambert

@ronan-a is in holidays for the next 2 weeks

However, I can answer: I'm currently using XOSTOR on a test cluster running on a fresh 8.3 and it works Can't tell for the upgrade though.

gb.123

@ronan-a

@ronan-a said in XOSTOR hyperconvergence preview:

@gb-123 What command and error exactly?

I was referring to the our earlier conversation :

@gb-123 said in XOSTOR hyperconvergence preview:

@ronan-a

running linstor command on both hosts gives the following error :

Traceback (most recent call last):
File "/usr/bin/linstor", line 21, in <module>
import linstor_client_main
ImportError: No module named linstor_client_main

@ronan-a said in XOSTOR hyperconvergence preview:

Regarding this error, can you say me what's your XCP-ng version? And what's the output of: yum list installed | grep -i linstor?

Output of yum list installed | grep -i linstor :
drbd.x86_64                     9.22.0-1.el7               @xcp-ng-linstor      
drbd-bash-completion.x86_64     9.22.0-1.el7               @xcp-ng-linstor      
drbd-pacemaker.x86_64           9.22.0-1.el7               @xcp-ng-linstor      
drbd-reactor.x86_64             1.0.0-1                    @xcp-ng-linstor      
drbd-udev.x86_64                9.22.0-1.el7               @xcp-ng-linstor      
drbd-utils.x86_64               9.22.0-1.el7               @xcp-ng-linstor      
drbd-xen.x86_64                 9.22.0-1.el7               @xcp-ng-linstor      
kmod-drbd.x86_64                9.2.2+ptf.1_4.19.0+1-1     @xcp-ng-linstor      
linstor-client.noarch           1.18.0-1                   @xcp-ng-linstor      
linstor-common.noarch           1.21.1-1.el7               @xcp-ng-linstor      
linstor-controller.noarch       1.21.1-1.el7               @xcp-ng-linstor      
linstor-satellite.noarch        1.21.1-1.el7               @xcp-ng-linstor      
python-linstor.noarch           1.18.0-1                   @xcp-ng-linstor      
xcp-ng-linstor.noarch           1.2-1.xcpng8.3             @xcp-ng-linstor      
xcp-ng-release-linstor.noarch   1.4-1.xcpng8.3             @xcp-ng-base
@ronan-a said in XOSTOR hyperconvergence preview:

Do you have a XCP-ng 8.3? And did you run the beta installation script on it?

I installed XCP-ng 8.3 beta from ISO and then applied all the patches that were available. To install linstor, I followed the 1st Post of this topic.

ronan-a

@gb-123 If you run LINSTOR on 8.3 (FYI, linstor packages on this version are considered unstable), you must call the linstor command like that:

> python3 /usr/bin/linstor

By default python2 is called and is not compatible with this version of LINSTOR:

> python2 /usr/bin/linstor resource list
Traceback (most recent call last):
  File "/usr/bin/linstor", line 21, in <module>
    import linstor_client_main
ImportError: No module named linstor_client_main

ronan-a

@Maelstrom96 We must update our documentation for that, This will probably require executing commands manually during an upgrade.

gb.123

@ronan-a

Are there any other packages in xcp-ng requiring python2 ?
Would it be advisable to change default python to python3 (by adding alias in bashrc) ?
Or do you think I should create a script which in-turn runs python3 command only for linstor ?

PS: By executing :

python3 /usr/bin/linstor

Result :

Error: Unable to connect to linstor://localhost:3370: [Errno 99] Cannot assign requested address

Although through XO, I can see the VM on Linstor SR running fine,

ronan-a

@gb-123 said in XOSTOR hyperconvergence preview:

Are there any other packages in xcp-ng requiring python2 ?

Probably yes for few packages.

Would it be advisable to change default python to python3 (by adding alias in bashrc) ?

I don't recommend to do that.

Or do you think I should create a script which in-turn runs python3 command only for linstor ?

It's probably the best solution for the moment. Our ideal goal would obviously to only have Python 3:

But currently XCP-ng 8.3 is still in development.
And there are still things to check before thinking about globally changing the default Python interpreter.

gb.123

@ronan-a

Any idea why I am getting :

Error: Unable to connect to linstor://localhost:3370: [Errno 99] Cannot assign requested address

I installed linstor with replication as 2 and the VM is running fine. I am executing this command from dom0 of Pool Master.

Swen

@gb-123 pool master does not need to be linstor controller. Try this command on the other hosts. It will only work on the node which is the linstor controller.

gb.123

@Swen

Thanks so much. It seems that the controller Is running on Node 2.

Btw, is there any way to control the node on which controller is installed (without loosing the data)? I remember installing on Node 1 first and then on Node 2. But it seems that Node 2 is actually running the controller.

I know the above is a noob question, but I just wanted to be sure if the linstor controller actually moves to another node when one is not available.

Maelstrom96

@ronan-a said in XOSTOR hyperconvergence preview:

@Maelstrom96 We must update our documentation for that, This will probably require executing commands manually during an upgrade.

Any news on that? We're still pretty much blocked until that's figured out.

Also, any news on when it will be officially released?

limezest

I have a 3 node cluster in my lab running 8.2.1 with thin provisioned LVM on SATA SSDs.

I changed the 10G management interfaces from being untagged to being being VLAN tagged and I did an emergency network reset following the changeover.
Each node only has one 10G interface which is used for all management, vm and storage traffic.

Following the network reset, the cluster did not automatically recover, so I restarted my three nodes.

Now the cluster is up and the nodes claim to be connected to the XOSTOR SR, but I cannot take snapshots, clone VMs, and some VMs fail to start with the following error:

SR_BACKEND_FAILURE_1200(, Empty dev path for 19f26a13-a09e-4c38-8219-b0b6b2d4dc26, but definition "seems" to exist, )

Any troubleshooting guidance is appreciated. Thanks in advance.

[edit1]
I am only able to run linstor commands such as 'linstor node list' on the node that is currently the linstor controller.

If I try linstor node list (or any linstor command) on the satellite nodes, i get the error:

Error: Unable to connect to linstor://localhost:3370: [Errno 99] Cannot assign requested address

linstor node interface list host1 gives me

host2       |   NetInterface     | IP            | Port   | EncryptionType
+ StltCon   |   default          | 10.10.10.11   | 3366   | PLAIN

I am able to use 'linstor-kv-tool' to find the linstor volume that maps to each of my virtual disk images, but only on the controller.

linstor-controller.service is running on host 2
linstror-satellite.service is running on hosts 0 and 1, but i don't see any process listening on 3366 from netstat -tulpn
[/edit1]