Intermediate Nodes are NOT working!!

Question

Level 1

4 points

Intermediate Nodes are NOT working!!

Hello,

I am trying to set up a mixed cluster of an intermediate node and a controller running Compressor. The controller will be a Mac with Compressor, Qmaster and Qadministrator installed; the intermediate node will be running Linux, which should be compatible since it is UNIX-based and has SSH.

After I did a fresh install of my Linux distro and installed the SSH server package, I went on Compressor 4.1 to add it as a remote intermediate node. It repeatedly spit out that either "The host is down" (it was never down) or "The host name could not be found for this address" (when I directly put in the local IP of my Linux machine).

I then gave up on Compressor 4.1 and its dumbed-down interface, and downgraded to Compressor 4.0.7. Under preferences it had the option of manually adding computers (to what, I don't know at all). Adding "localhost" came up with the right IP address and the status said "Online" but adding the Linux machine's hostname "server.local" resulted in the right IP address, but the status always says "Not sharing." At least it can identify the Linux server now.

Then I followed Apple's documentation and went to the Qmaster menu and selected the Share This Computer menu. Here I could select which kind of cluster I want to create; I selected QuickCluster With Services, with the cluster name "Home render farm." In Rendering Options I could add the Linux machine by SSH; this was authenticated, and it said "Host authenticated." In any case the cluster, even if the Linux machine wasn't working, should be working.

But under Qadministrator, NO CLUSTER SHOWS UP AT ALL! The services for localhost show up as unmanaged, but I am unable to either create a cluster with the Mac as the Cluster Controller or see my previously made cluster "Home rendering farm" at all.

Occasionally the Home rendering farm cluster does show up, but I can't do anything with it. It either fails, disappears or cannot be identified...

Can anyone help?

MacBook Pro (15-inch Early 2011), OS X Mavericks (10.9), 2011 15" 2.2Ghz Quad Core, 16GB RAM

Posted on Jan 12, 2014 8:34 PM

Reply

Answer 1

Jan 15, 2014 2:55 AM in response to Prodo123

HI , this is a common problem with clusters in Qmaster in the legacy FCS7, FCPX and the prior Compressor V4.0 where the clusters wont show up or they are not communicating.

Unfortunately it's a bit of black magic and science and Hong Kong fortune-teller to get up and working.😉

I can assume your very knowledgeable/experienced and forgive me for confirming:

and all your network, DNS etc its working 100% (using OSX server?)
file sharing ACEs/ACL's are ok for the file system paths for the nodes? (worth a look although not you cause here)
preferred network paths are working ....

However, much of the  Apple Qmaster inoperability is due to old state information. Usually and I believe simply, is to REMOVE (purge | delete) the Apple Qmaster state information and fire it up again. Try this procedure....

(I'm doing this from memory as I've upgraded to FCPX 10.1 and Compressor V4.1 .. its changed))

A) Start here:

in Compressor.app / Apple Qmaster/ V4.0 (or Apple Qadministrator.app on all Nodes), record your cluster settings and network interface paths.
in Compressor V4.0 / Apple Qmaster/ , stop sharing on all the nodes. (tick off share this.. etc )
view the OSX Activity Monitor (or ps ) to determine tha ALL the compressor and rendering tasks have stopped. Do this on all nodes.
quit compressor on all nodes.
delete the directory (folder) at ~/Library/Application Support/Apple Qmaster/!.. yep! delete it or use finder to move to rubbish (trash). Apple Qmaster components will allocate and populate it again when you crack up Apple Qmaster Sharing (tick tha tick on in Step 2)
oh Warning here: DONT DELETE ANY COMPRESSOR.app V4.0 PReference and state informations - these have nothing to do wth Qmaster. (~/library/application support/compressor/
take a breath....

B) Re-Establishing Qmaster Cluster. - use a MANAGED CLUSTER not a quickcluster.

check all your network and file permissions on all your source and destination nodes. just a last check.
assume you are NOT using a quickcluster. Use a managed cluster.
Launch Compressor.app V4.0 your controller node.
Start Compressor.app menus/ Apple Qmaster ( enable share this computer as u normally would). All defaults are in place - care here
I would suggest you reduce the network issues to nodes for communication and FILE SYSTEM sharing by
- specifying a DEDICATED network path to all nodes - all use EN3 so that you can diagnose the Network traffic over the node to other nodes.
- re-iinstate your cluster storage if you used it: "shared storage" else let it default to your ~/library/Application support/Apple Qmaster/Storage
- NOTE: if you ruse a dedicate CLUSTER STORAGE , make sure that FILE SYSTEM is mounted and available to your Qmaster controller and its services and to any other nodes. Obvious, however worth mentioning. Oh and while you're there, please just check the permissions of the folder .
change the instances you need as usual for rendering or transcoding.
tick use MANAGED SERVICES only.
Launch Compressor on one of your SERVICES node hosts and perform steps 3-5 on that node with the exception of the controller. (Services only). Im sure you know this.
test the network connection. use traceroute or ping with the -I (iface) parm to direct it down the specific dedicated NIC you're using for step 4. above. Assume all is well
use Apple Qadministrator.app is you like to set up your cluster, else just use what you normally do.

C) Verifying you cluster services and cluster storage. - this is where you will see if its all working.

make sure ALL the instances of the Qmaster subtasks are running. Here's how:
- for the Controller and Services node and other nodes: doe a GEAR DOWN (RMButton / cntl+click) on teh COMPRESSOR.app V4.0 icon in the dock and you will see a list of ACTIVE SERVICES running.
- This is SHOULD be equal to the number of instances you specified. (nothing there for Compressor 4.1 BTW (dec 2013).
- IF some are captured or some are not started, just be patient. Sometimes this can takes minutes.
- view also in the ACTIVITY MONITOR.app and also Apple Qadministrator.app.. you will see al the subtasks ready.
now.. MOUNT THE cluster storage on each node and the cluster controller. THIS IS KEY!
- use Compressor.app menu / Mount cluster Storage)
- immediately you will be presented with a LISTBOX and hopefully the shared name of your cluster storage will be in the LIST BOX. Select it and you will have the cluster storage mounted on your desktop. Your Apple Qmaster is available from this node.
- if it's NOT then and your list box is UNPOPULATED, your cluster WONT WORK. It's symptomatic or the Apple Qmaster services not started or corrupt.
  if this is the case, start at STEP A again and redo the whole procedure... its fast if you have done it 2-3 times.😀
do for all your service nodes.
Verify the cluster with Apple Qadministrator.app. All should be well!

D) Submit a simple batch job and monitor the transaction using SHARE MONITOR.app

set up your render scripts again or transcode job.also check the Apple Qmaster logs in ~/library/application support/Apple Qmaster/Logs for each service node , the JOB CONTROLLER and the services controller logs for errors using the console.app

I think this enough to get this working. It's worked many times and demonstrates that Compressor.app V4.0 and Apple Qmaster are sound applications that are subject to stale and corrupt state information that can be caused by abnormal shutdowns and network outages.

I Hope that helps.

Please post your results for others to see.

Warwick

Hong Kong

Certified  ACTC

Reply