XOSTOR hyperconvergence preview

ronan-a

@Maelstrom96 said in XOSTOR hyperconvergence preview:

The folder /dev/drbd/by-res/ doesn't exist currently.

You're lucky, I just produced a fix yesterday to fix this kind of problem on pools with more than 3 machines: https://github.com/xcp-ng/sm/commit/f916647f44223206b24cf70d099637882c53fee8

Unfortunately, I can't release a new version right away, but I think this change can be applied to your pool.
In the worst case I'll see if I can release a new version without all the fixes in progress...

0

Wescoeur committed to xcp-ng/sm

fix(LinstorSR): ensure vhdutil calls are correctly executed on pools with > 3 hosts

Signed-off-by: Ronan Abhamon <ronan.abhamon@vates.fr>

Maelstrom96

@ronan-a said in XOSTOR hyperconvergence preview:

You're lucky, I just produced a fix yesterday to fix this kind of problem on pools with more than 3 machines: https://github.com/xcp-ng/sm/commit/f916647f44223206b24cf70d099637882c53fee8

Unfortunately, I can't release a new version right away, but I think this change can be applied to your pool.
In the worst case I'll see if I can release a new version without all the fixes in progress...

Thanks, that does look like it would fix the missing drbd/by-res/ volumes.

Do you have an idea about the missing StoragePool for the new host that was added using linstor-manager.addHost? I've checked the code and it seems like it might just provision the SP on sr.create?

Also, I'm not sure how feasible it would be for SM but having a nightly-style build process for those cases seems like it would be really useful for hotfix testing.

drfill

Hello guys,
Awesome realisation, works like a charm! But security in low level. Anyone who wants break down disk cluster/HC storage does it. I installed and as I see Linstor Controller ports opened to whole world. Any solution have you to close external port (when management in global IP), and communicate through Storage Network?

olivierlambert

Hmm in theory I would say it should only listen on the management network (or the storage network, but not everything)

Maelstrom96

@ronan-a Any news on when the new version of linstor SM will be released? We're actually hard blocked by the behavior with 4 nodes right now so we can't move forward with a lot of other tests we want to do.

We also worked on doing a custom build of linstor-controller and linstor-satellite to allow support of Centos 7 with it's lack of sedsid -w support, and we might want to see if we could get a satisfactory PR that could be merged into linstor-server master so that people using XCP-ng can also use linstor's built-in snapshot shipping. Since K8s linstor snapshotter uses that functionality to provide volume backups, it means that using K8s with linstor on XCP-ng is not really possible unless this is fixed.

Would that be something that you guys could help us push to linstor?

ronan-a

@Maelstrom96

Do you have an idea about the missing StoragePool for the new host that was added using linstor-manager.addHost? I've checked the code and it seems like it might just provision the SP on sr.create?

If I remember correctly, this script is only here for adding the PBDs of the new host and configuring the services. If you want to add a new device, it is necessary to manually create a new VG LVM, and add it via a linstor command.

Also, I'm not sure how feasible it would be for SM but having a nightly-style build process for those cases seems like it would be really useful for hotfix testing.

The branch was in a bad state (many fixes to test, regressions, etc). I was able to clean all that up, it should be easier to do releases now.

Any news on when the new version of linstor SM will be released?

Today.

ronan-a

WARNING: I just pushed new packages (they should be available in our repo in few minutes) and I made an important change in the driver which requires manual intervention.

minidrbdcluster is no longer used to start the controller, instead we use drbd-reactor which is more robust.
To update properly, you must:

Disable minidrbdcluster on each host: systemctl disable --now minidrbdcluster.
Install new LINSTOR packages using yum:

blktap-3.37.4-1.0.1.0.linstor.1.xcpng8.2.x86_64.rpm
xcp-ng-linstor-1.0-1.xcpng8.2.noarch.rpm
xcp-ng-release-linstor-1.2-1.xcpng8.2.noarch.rpm
http-nbd-transfer-1.2.0-1.xcpng8.2.x86_64.rpm
sm-2.30.7-1.3.0.linstor.7.xcpng8.2.x86_64.rpm

On each host, edit /etc/drbd-reactor.d/sm-linstor.toml. (Note: it's probably necessary to create the folder /etc/drbd-reactor.d/ with mkdir.) And add this content:

[[promoter]]

[promoter.resources.xcp-persistent-database]
start = [ "var-lib-linstor.service", "linstor-controller.service" ]

After that you can manually start drbd-reactor on each machine: systemctl enable --now drbd-reactor.

You can reuse your SR again.

Swen

@ronan-a Just to be sure: IF you install it from scratch you can still use the installation instructions from the top of this thread, correct?

ronan-a

@Swen Yes you can still use the installation script. I just changed a line to install the new blktap, so redownload it if necessary.

Swen

@ronan-a perfect, thx! Is this the new release Olivier was talking about? Can you provide some information when to expect the first stable release?

fred974

@ronan-a Just to be clear on what I have to do..

Diable minidrbdcluster on each host: systemctl disable --now minidrbdcluster no issue here...
Install new LINSTOR packages. How do we do that? Do we run the installer again by running:
wget https://gist.githubusercontent.com/Wescoeur/7bb568c0e09e796710b0ea966882fcac/raw/1707fbcfac22e662c2b80c14762f2c7d937e677c/gistfile1.txt -O install && chmod +x install
or
./install update

Or do I simply installed the new RPM without running the installer?

wget --no-check-certificate blktap-3.37.4-1.0.1.0.linstor.1.xcpng8.2.x86_64.rpm
wget --no-check-certificate xcp-ng-linstor-1.0-1.xcpng8.2.noarch.rpm
wget --no-check-certificate xcp-ng-release-linstor-1.2-1.xcpng8.2.noarch.rpm
wget --no-check-certificate http-nbd-transfer-1.2.0-1.xcpng8.2.x86_64.rpm
wget --no-check-certificate sm-2.30.7-1.3.0.linstor.7.xcpng8.2.x86_64.rpm
yum install *.rpm

Where do we get this files from? what is the URL?

On each host, edit /etc/drbd-reactor.d/sm-linstor.toml no problem here...

Can you please confirm which procedure to use for steps 2?

Thank you.

fred974

@ronan-a is that the correct URL?

https://koji.xcp-ng.org/kojifiles/packages/blktap/3.37.4/1.0.1.0.linstor.1.xcpng8.2/x86_64/blktap-3.37.4-1.0.1.0.linstor.1.xcpng8.2.x86_64.rpm

https://koji.xcp-ng.org/kojifiles/packages/xcp-ng-linstor/1.0/1.xcpng8.2/noarch/xcp-ng-linstor-1.0-1.xcpng8.2.noarch.rpm

https://koji.xcp-ng.org/kojifiles/packages/xcp-ng-release-linstor/1.2/1.xcpng8.2/noarch/xcp-ng-release-linstor-1.2-1.xcpng8.2.noarch.rpm

https://koji.xcp-ng.org/kojifiles/packages/http-nbd-transfer/1.2.0/1.xcpng8.2/x86_64/http-nbd-transfer-1.2.0-1.xcpng8.2.x86_64.rpm

https://koji.xcp-ng.org/kojifiles/packages/sm/2.30.7/1.3.0.linstor.7.xcpng8.2/x86_64/sm-2.30.7-1.3.0.linstor.7.xcpng8.2.x86_64.rpm

ronan-a

@Swen said in XOSTOR hyperconvergence preview:

@ronan-a perfect, thx! Is this the new release Olivier was talking about? Can you provide some information when to expect the first stable release?

If we don't have a new critical bug, normally in few weeks.

@fred974

Or do I simply installed the new RPM without running the installer?

You can update the packages just using yum if you already have the xcp-ng-linstor yum repo config. There is no reason to download manually the packages from koji.

fred974

@ronan-a ok,
I managed to install using yum

yum install blktap.x86_64
yum install xcp-ng-linstor.noarch
yum install xcp-ng-release-linstor.noarch
yum install http-nbd-transfer.x86_64

But I cannot find sm-2.30.7-1.3.0.linstor.7.xcpng8.2.x86_64.rpm.
Is it yum install sm-core-libs.noarch ?

ronan-a

@fred974 What's the output of yum update sm?

fred974

systemctl enable --now drbd-reactor

Job for drbd-reactor.service failed because the control process exited with error code. See "systemctl status drbd-reactor.service" and "journalctl -xe" for details.

systemctl status drbd-reactor.service

[12:21 uk xostortmp]# systemctl status drbd-reactor.service
* drbd-reactor.service - DRBD-Reactor Service
   Loaded: loaded (/usr/lib/systemd/system/drbd-reactor.service; enabled; vendor preset: disabled)
   Active: failed (Result: exit-code) since Thu 2023-03-23 12:12:33 GMT; 9min ago
     Docs: man:drbd-reactor
           man:drbd-reactorctl
           man:drbd-reactor.toml
  Process: 8201 ExecStart=/usr/sbin/drbd-reactor (code=exited, status=1/FAILURE)
 Main PID: 8201 (code=exited, status=1/FAILURE)

journalctl -xe has no usefull information but the SMlog log file has the following:

Mar 23 12:29:27 uk SM: [17122] Raising exception [47, The SR is not available [opterr=Unable to find controller uri...]] Mar 23 12:29:27 uk SM: [17122] lock: released /var/lock/sm/a20ee08c-40d0-9818-084f-282bbca1f217/sr Mar 23 12:29:27 uk SM: [17122] ***** generic exception: sr_scan: EXCEPTION <class 'SR.SROSError'>, The SR is not available [opterr=Unable to find controller uri...] Mar 23 12:29:27 uk SM: [17122] File "/opt/xensource/sm/SRCommand.py", line 110, in run Mar 23 12:29:27 uk SM: [17122] return self._run_locked(sr) Mar 23 12:29:27 uk SM: [17122] File "/opt/xensource/sm/SRCommand.py", line 159, in _run_locked Mar 23 12:29:27 uk SM: [17122] rv = self._run(sr, target) Mar 23 12:29:27 uk SM: [17122] File "/opt/xensource/sm/SRCommand.py", line 364, in _run Mar 23 12:29:27 uk SM: [17122] return sr.scan(self.params['sr_uuid']) Mar 23 12:29:27 uk SM: [17122] File "/opt/xensource/sm/LinstorSR", line 634, in wrap Mar 23 12:29:27 uk SM: [17122] return load(self, *args, **kwargs) Mar 23 12:29:27 uk SM: [17122] File "/opt/xensource/sm/LinstorSR", line 560, in load Mar 23 12:29:27 uk SM: [17122] raise xs_errors.XenError('SRUnavailable', opterr=str(e)) Mar 23 12:29:27 uk SM: [17122] Mar 23 12:29:27 uk SM: [17122] ***** LINSTOR resources on XCP-ng: EXCEPTION <class 'SR.SROSError'>, The SR is not available [opterr=Unable to find controller uri...] Mar 23 12:29:27 uk SM: [17122] File "/opt/xensource/sm/SRCommand.py", line 378, in run Mar 23 12:29:27 uk SM: [17122] ret = cmd.run(sr) Mar 23 12:29:27 uk SM: [17122] File "/opt/xensource/sm/SRCommand.py", line 110, in run Mar 23 12:29:27 uk SM: [17122] return self._run_locked(sr) Mar 23 12:29:27 uk SM: [17122] File "/opt/xensource/sm/SRCommand.py", line 159, in _run_locked Mar 23 12:29:27 uk SM: [17122] rv = self._run(sr, target) Mar 23 12:29:27 uk SM: [17122] File "/opt/xensource/sm/SRCommand.py", line 364, in _run Mar 23 12:29:27 uk SM: [17122] return sr.scan(self.params['sr_uuid']) Mar 23 12:29:27 uk SM: [17122] File "/opt/xensource/sm/LinstorSR", line 634, in wrap Mar 23 12:29:27 uk SM: [17122] return load(self, *args, **kwargs) Mar 23 12:29:27 uk SM: [17122] File "/opt/xensource/sm/LinstorSR", line 560, in load Mar 23 12:29:27 uk SM: [17122] raise xs_errors.XenError('SRUnavailable', opterr=str(e))

Is it normal that the XOSTOR SR is still visible in XO?

fred974

@ronan-a said in XOSTOR hyperconvergence preview:

What's the output of yum update sm?

[12:33 uk ~]# yum update sm

Loaded plugins: fastestmirror
Loading mirror speeds from cached hostfile
Excluding mirror: updates.xcp-ng.org
 * xcp-ng-base: mirrors.xcp-ng.org
Excluding mirror: updates.xcp-ng.org
 * xcp-ng-linstor: mirrors.xcp-ng.org
Excluding mirror: updates.xcp-ng.org
 * xcp-ng-updates: mirrors.xcp-ng.org
No packages marked for update

ronan-a

@fred974 And rpm -qa | grep sm? Because the sm LINSTOR package update is in our repo. So I suppose you already installed it using koji URLs.

fred974

@ronan-a said in XOSTOR hyperconvergence preview:

@fred974 And sudo rpm -qa | grep sm? Because the sm LINSTOR package update is in our repo. So I suppose you already installed it using koji URLs.

microsemi-smartpqi-1.2.10_025-2.xcpng8.2.x86_64
smartmontools-6.5-1.el7.x86_64
sm-rawhba-2.30.7-1.3.0.linstor.7.xcpng8.2.x86_64
ssmtp-2.64-14.el7.x86_64
sm-cli-0.23.0-7.xcpng8.2.x86_64
sm-2.30.7-1.3.0.linstor.7.xcpng8.2.x86_64
libsmbclient-4.10.16-15.el7_9.x86_64
psmisc-22.20-15.el7.x86_64

Yes, I installed it from the koji URLs before seeing your rely

ronan-a

@fred974 I just repaired your pool, there was a small error in the conf that I gave in my previous post.