[SOLVED] Support Storage Type - Quobyte (Client)

[DerDanilo]

@proxmox Team @tom

I directly opened a bugzill enhancement report to get this rolling as soon as possible. Should be relatively easy to implement and this could even be available within Q1 2019.

https://bugzilla.proxmox.com/show_bug.cgi?id=2047

Thanks guys!

 

[DerDanilo]

I tried to adjust the GlusterFSPlugin into a custom storage plugin for Quobyte today.

The issue I'm currently having is:

Error loading storage plugin "PVE::Storage::Custom::QuobytePlugin": does not provide an api() method

When adding the following to the file (found in a forum)...

# API version
sub api {
    return 1;
}

... it fails with compilation errors:

Jan 04 18:11:37 node4 pvedaemon[12734]: Compilation failed in require at /usr/share/perl5/PVE/API2.pm line 13, <DATA> line 755.
Jan 04 18:11:37 node4 pvedaemon[12734]: BEGIN failed--compilation aborted at /usr/share/perl5/PVE/API2.pm line 13, <DATA> line 755.
Jan 04 18:11:37 node4 pvedaemon[12734]: Compilation failed in require at /usr/share/perl5/PVE/Service/pvedaemon.pm line 8, <DATA> line 755.
Jan 04 18:11:37 node4 pvedaemon[12734]: BEGIN failed--compilation aborted at /usr/share/perl5/PVE/Service/pvedaemon.pm line 8, <DATA> line 755.
Jan 04 18:11:37 node4 pvedaemon[12734]: Compilation failed in require at /usr/bin/pvedaemon line 11, <DATA> line 755.
Jan 04 18:11:37 node4 pvedaemon[12734]: BEGIN failed--compilation aborted at /usr/bin/pvedaemon line 11, <DATA> line 755.
Jan 04 18:11:37 node4 systemd[1]: pvedaemon.service: Control process exited, code=exited status=255
Jan 04 18:11:37 node4 systemd[1]: Failed to start PVE API Daemon.
Jan 04 18:11:37 node4 systemd[1]: pvedaemon.service: Unit entered failed state.
Jan 04 18:11:37 node4 systemd[1]: pvedaemon.service: Failed with result 'exit-code'.

Any ideas what I am missing?
Any tips?

The current state of the plugin is available on Github:
https://github.com/DerDanilo/proxmox-stuff/blob/master/quobytestorageplugin/QuobytePlugin.pm

 

[Stoiko Ivanov]

hmm - seems not the complete error-trace is present in the quoted error-messages.

quick guess: did you include the 'sub api ..' part in the bottom of your module (i.e. .pm-file)? If so you might have put it after the `1;` line in the bottom - and a perl-module should end in a `true` statement - see e.g. https://stackoverflow.com/questions/5293246/why-the-1-at-the-end-of-each-perl-package

 

[DerDanilo]

Yes the sub api is configured. It looks just like in other examples for custom storage plugins I found on Github.

https://github.com/DerDanilo/proxmo...0da/quobytestorageplugin/QuobytePlugin.pm#L42

I have no clue what to put for the API version. Either it produces the error above or it states that the version is "too old".

Can you please check the code on github and see if there are any issues? I am no developer and just copied and adjusted the GlusterFSPlugin, since Quobyte works quiet similiar to GlusterFS.

 

[wolfgang]

Hi DerDanilo,

Quobyte is closed source and so it will not come in Proxmox VE.
We can and will only support opensource technologies.

Anyway, it locks like Quobyte support NFS natively what is wrong with this approach?

 

[dcsapak]

it seems that you defined properties that are already defined, you can see this in the complete journal output:

Jan 08 10:07:34 pve-ceph-01 pveproxy[12251]: Plugin "PVE::Storage::Custom::QuobytePlugin" is implementing an older storage API, an upgrade is recommended
Jan 08 10:07:34 pve-ceph-01 pveproxy[12251]: duplicate property 'server' at /usr/share/perl5/PVE/SectionConfig.pm line 176, <DATA> line 755.
Jan 08 10:07:34 pve-ceph-01 pveproxy[12251]: Compilation failed in require at /usr/share/perl5/PVE/API2/Cluster.pm line 13, <DATA> line 755.

since the storage config is a 'section config' we can define each property only one time

 

[DerDanilo]

Hi @wolfgang

Quobyte is closed source and so it will not come in Proxmox VE.
We can and will only support opensource technologies.

I know and I understood that. But I asked for a hint on what is wrong with the sub api and not to solve the code generally, since I try to understand the code somehow. ;-)
Update: @dcsapak Thanks!

NFS natively what is wrong with this approach

Single point of failure?! To my knowledge NFS does not support multiple backends to continue active streams if one device fails.

---

But since we won't be getting anywhere here I created another post that could generally solve issues like this one here.

https://forum.proxmox.com/threads/directory-storage-type-issues.50457/

Thanks!

 

[DerDanilo]

Using the Directory Storage Plugin and the options "is_mountpoint 1" and "mkdir 0" the issue could be solved. No additional plugin needed. Maybe Quobyte will implement a Plugin in future but it works just fine with the mountpoint.

Thanks for your help guys!