Daniel_OM's dashboard

Yes, I'm glad we reached at the same conclusion ;)
At the moment it's not possible to use both QMaster Controller and MatrixStore on a node. As I wrote few posts above - maybe one day we
will support that but it's rather unlikely because we would not have control over machine anymore, we wouldn't know how processes interact, what is the CPU load during encoding etc.
So basically it can be a shot in a foot

My advice:
If you have a spare machine usu it as a QMaster controller and the rest of the nodes as clients. MatrixStore should work fine on them I guess.
If you don't have a spare machine uninstall MatrixStore from all of nodes and install it only on two - two node cluster.

About error 301.
That's reallly strange as Dropspot and plugin should work fine with default user. I don't know why cluster complains about it. Maybe cluster was not installed properly (QMaster) and complains because of security problems on default users. Did you send log files on failled and succesful connection to the vault?

About IPs in the vault file:
That's a bug in AdminTool which has been fixed some time ago. I believe if you download the latest version of MatrixStore software it should be fine.
In the previous versions you have to add IPs manually - as you did.

Thanks for nice words

Cheers

Don't worry about port 32121, it is used only for debug purposes, the output is fine. Netstat also looks good.

The problem with Dropspot is weird though. It might be the case that something went wrong during installation.

Could you please try to create a new vault using Admin and try to connect with Dropspot to it?
Please also send log files from all machines.

Reinstallation is the last step, better to know what happened :)

Unfortunatelly we won't be able to reply tomorrow, so if you want to install a new cluster - feel free, but please make sure you send us logs first

Regards
Daniel

Yes, by using only service and not controller on node41 MatrixStore should work fine.
The output of netstat looks better. What's the state of the cluster in AdminTool. How's Dropspot doing?

I think only the node41 (controller) runs NFS service.
The other two are clients and use NFS share therefore they are not running portmapper (which is the source of the problems)
RPC errors appear because both NFS and MatrixStore are using RPC calls and both services want to launch portmapper.

Before uninstalling QMaster, could you launch netstat command on two other nodes and post the output?

Hi Lucas

The problem, as you pointed out, is that Qmaster uses NFS sharing.
So even if you disable the NFS it still will be started by Qmaster service.

We don't support additional services on MatrixStore nodes because of such reasons.

So advice for now is:
Try to switch of QMaster on that machine temporarily, just to check if that's the source of the problems.
If yes, maybe wipe the MatrixStore cluster on all nodes completely and install it again but only on two machines which don't use Qmaster.

The MatrixStore will be fine with only two nodes, you just won't have additional protection and benefits of having third node (data regeneration).

Maybe at some point we will release a new server compatible with NFS sharing.

Regards

If you disable nfsd by:

sudo nfsd disable

it should work i.e. the service will be switched off permanently. Please reboot after that and send the mxs log file from that machine and he output of netstat -a | grep LISTEN again

Message

rpc.statd[29] Failed to contact host macpro-8core4: RPC: Unknown host

is not related to MatrixStore. It seems to be the problem with NFS setup.
To me it looks like server or client cannot resolve DNS names.
Try to use IPs instead when setting up NFS share.

Hi

As I wrote before

You need to switch off the NFS sharing using Server Admin.
After that you HAVE TO REBOOT machine

Killing the process is not enough because:
- it stops only the NFS daemon and not other services (i.e. portmapper)
- daemon will restart after reboot

D.

This issue is / will be solved in the version 1.1.9 of the plugins

Hi Lucas

Stop NFS sharing on node 41

Please note that switching off sharing using Server Admin does NOT turn off the service completely. You need to reboot the machine after doing that.

If that doesn't help please send output of a command:

netstat -a | grep LISTEN

launched on that machine.

Thanks

Hi Lucas
I got your message. Thanks for the details.
We will post a message here or sent you an email as soon as we will reproduce the problem.

Stay tuned

We recommend to install FinalCutServer on a dedicated machine, not a MatrixStore node. FCS is meant to be used by multiple users and for that reason (and because all its actions consume a lot of CPU and memory) we normally install FCS on a separate machine.

Our MatrixStoreConnect software (FTP frontend for FCS for example) can be installed on any machine, works particularly well installed on FCS machine.

The concept of Head node is our intarnal term. End user don't know which machine at the moment is considered as main. All nodes in a cluster are equal :)

Yes, you can try to use MatrixStore nodes as Qmaster cluster. Actually is could be a good way to utilize machines when MatrixStore is not used. We do not recommend that though as we did not test that.
Encoding is really expensive operation - may happen that node may run out of memory, stop respond etc...
Besides for security reasons we disable some services from OSX during MatrixStore node instalation. If QMaster relies on them it may not work...
But as a start it may be a good idea to use for example only 2 out of three MatrixStore nodes as QMaster machines. You are welcome to test that...

You cannot ingest directly to MatrixStore. It is used as an archive device so prefered way of working is put your data on local disk, SAN etc.. edit it and archive using Dropspot or FinalCutServer. The last one can automatically delete the local copy so you don't have to do this manually.

Hope that helps.