Re: [HTCondor-users] Unable to correctly create multi-machine pool

[Todd Tannenbaum <tannenba@xxxxxxxxxxx>]

On 4/24/2024 1:51 PM, Thurimella, Vijay wrote:

Hello,

I am trying to run a this Pegasus workflow for an experiment I am running. In order to run the workflow, I was trying to create a multi-machine condor pool using the instructions in the documentation from here. Whenever I run through the commands on the webpage and get to the point where I run condor_status on the submit node. I am getting the following error. 

Error: communication error

SECMAN:2007:Failed to end classad message.  

I am very new to HTCondor so any advice to help me get my multi machine pool running would be greatly appreciated.

I am creating this multi-machine pool using cloud lab. Each node is a m510 machine running ubuntu 22.04.02 LTS. The machines are all connected to the same network and each node has a hostname node{num}. I made node0 the central manager, node1 the submit node, and node2/node3 execute nodes. The commands I ran to create the multi-machine pool were:

$ curl -fsSL https://get.htcondor.org | sudo GET_HTCONDOR_PASSWORD="$htcondor_password" /bin/bash -s -- --no-dry-run --central-manager node0

$ curl -fsSL https://get.htcondor.org | sudo GET_HTCONDOR_PASSWORD="$htcondor_password" /bin/bash -s -- --no-dry-run --submit node0

$curl -fsSL https://get.htcondor.org | sudo GET_HTCONDOR_PASSWORD="$htcondor_password" /bin/bash -s -- --no-dry-run --execute node0

Thanks,

Vijay

    Hello Vijay, and welcome to the HTCondor community!
    
    Some quick thoughts :
    
    Re the commands you ran above, just want to confirm that you
    replaced "$htcondor_password" with an actual password you want to
    use to secure your pool, correct?  If the above is literally what
    you entered, I am guessing the shell would expand the password into
    an empty string.  
    
    Another thought... you mentioned the nodes are all connected to the
    same network, but perhaps the nodes are running a firewall?  By
    default HTCondor uses network port 9618, so that will need to be
    open.
    
    Also, do your machines have /etc/hosts or DNS setup such that, for
    instance, on node1 you can resolve hostname node0 ?  I.e. while
    logged into node1, can you do "ping node0" ?
    
    Finally, this may not matter, but personally I have always used
    fully-qualified hostnames to specify the central manager.  In other
    words, on the lines above, instead of node0, perhaps
    "node0.mydomain.org" (whatever your domain is).  Besides
    fully-qualified hostnames, I have also used IP addresses.
    
    Hope the above helps.  If you are still having trouble, please drop
    another note.
    
    regards,
    Todd