WIP to add lammps

README.md

@@ -191,11 +191,11 @@ We could also try a container - I will likely do both.
 
 ## TODO
 
+Grab one of these if you'd like to contribute!
+
 - A non-interactive submit needs to write a submission file (might not be needed)
 - Tweak the configs to get envars that have the cluster size (right now just hard coded to 2)
 - Test out [containers](https://chtc.cs.wisc.edu/uw-research-computing/docker-jobs)
-- LAMMPS example working
-
 
 ## License
 

config/manager/kustomization.yaml

@@ -5,4 +5,4 @@ kind: Kustomization
 images:
 - name: controller
   newName: ghcr.io/converged-computing/htcondor-operator
-  newTag: latest
+  newTag: test

controllers/htcondor/htcondor.go

@@ -164,6 +164,14 @@ func (r *HTCondorReconciler) getConfigMap(
 		data["start-manager"] = managerStart
 		data["start-execute"] = executeStart
 		data["start-submit"] = submitStart
+
+		// This will be copied into
+		// /root/secrets/token
+		//token, err := r.getToken(ctx, cluster)
+		//if err != nil || token == "" {
+		//	return cm, ctrl.Result{Requeue: true}, err
+		//}
+		//data[tokenKey] = token
 	}
 
 	// Create the config map with respective data!

controllers/htcondor/templates.go

@@ -32,6 +32,9 @@ var startSubmitTemplate string
 //go:embed templates/components.sh
 var startComponents string
 
+//go:embed templates/token.sh
+var tokenTemplate string
+
 // NodeTemplate populates a node entrypoint
 type NodeTemplate struct {
 	Node        api.Node

controllers/htcondor/templates/components.sh

@@ -7,6 +7,10 @@
 echo "Sleeping for networking..."
 sleep 3
 
+# Copy over our token, and remove newline at top
+# mkdir -p /root/secrets
+# tr --delete '\n' < /htcondor_operator/token > /root/secrets/token
+
 # Initialization commands
 {{ .Node.Commands.Init}} > /dev/null 2>&1
 
@@ -20,17 +24,35 @@ mkdir -p ${workdir}
 {{end}}
 
 {{define "config"}}
+
+# This won't trigger to the right set of conditions in 
+# /etc/condor/config.d/01-security.conf if we set to yes
+# with yes, the basic cluster works, but parallel universe does not
+export USE_POOL_PASSWORD=yes
+
 # Shared logic to write a config across nodes
 echo "NEGOTIATOR_INTERVAL=10" >> /etc/condor/condor_config.local
-# echo "SEC_PASSWORD_FILE = $(LOCAL_DIR)/lib/condor/pool_password" >> /etc/condor/condor_config.local
 
-# Generate password
-mkdir -p /root/secrets
-# chmod 0700 /root/secrets
+# note that this container setup seems to already be using a USE_POOL_PASSWORD with host auth
+# These don't seem to work for this setup
+# echo "use security:host_based" >> /etc/condor/config.d/00-insecure.config
+
 # The top one is shown for docker-compose, the second in the config example
+mkdir -p /root/secrets
 condor_store_cred -p {{.Spec.Config.Password}} -f /root/secrets/pool_password 
 # condor_store_cred -p {{.Spec.Config.Password}} -f /var/lib/condor/pool_password
 
+# Try updating to allow TOKEN and PASSWORD
+echo "SEC_DEFAULT_AUTHENTICATION_METHODS = FS, PASSWORD, TOKEN, IDTOKENS" >> /etc/condor/config.d/01-security.conf 
+echo "ALLOW_WRITE = *" >> /etc/condor/config.d/01-security.conf 
+
+# Austin DANGER POWERS!
+echo 'ALLOW_ADVERTISE_STARTD = condor_pool@*/* $(ALLOW_ADVERTISE_STARTD)' >> /etc/condor/config.d/01-security.conf 
+echo 'ALLOW_ADVERTISE_SCHEDD = condor_pool@*/* $(ALLOW_ADVERTISE_SCHEDD)' >> /etc/condor/config.d/01-security.conf 
+echo 'ALLOW_ADVERTISE_MASTER = condor_pool@*/* $(ALLOW_ADVERTISE_MASTER)' >> /etc/condor/config.d/01-security.conf 
+
+# ALLOW_WRITE = *
+
 # TODO this should be actual cpus, not nodes
 export NUM_CPUS={{.Spec.Size}}
 {{end}}
@@ -40,11 +62,61 @@ export NUM_CPUS={{.Spec.Size}}
 {{ end }}
 
 {{define "condor-host"}}
-
-export USE_POOL_PASSWORD=yes
 export CONDOR_HOST={{ .ClusterName }}-manager-0-0.{{ .Spec.ServiceName }}.{{ .Namespace }}.svc.cluster.local
 # export CONDOR_SERVICE_HOST=${CONDOR_HOST}
 {{ end }}
 
+{{define "approve-tokens"}}
+# This snippet will always approve token requests, if needed
+# Install jq for now - a hack to approve token requests
+yum install -y jq
+
+# Ideally we can provide this via a config or the condor_token_request_auto_approve that takes a hostname
+while true
+do
+    for requestid in $(condor_token_request_list -json | jq -r .[].RequestId); do
+        echo "yes" | condor_token_request_approve -reqid ${requestid}
+    done
+    sleep 15
+done
+{{end}}
+
+{{define "security-config"}}
+tee -a /etc/condor/config.d/01-security.conf <<EOF
+# Require authentication and integrity checking by default.
+use SECURITY : With_Authentication
+
+# Host-based security is fine in a container environment, especially if
+# we're also using a pool password or a token.
+use SECURITY : Host_Based
+# We also want root to be able to do reconfigs, restarts, etc.
+ALLOW_ADMINISTRATOR = root@$(FULL_HOSTNAME) condor@$(FULL_HOSTNAME) $(ALLOW_ADMINISTRATOR)
+
+# SEC_DEFAULT_AUTHENTICATION_METHODS = FS, PASSWORD, TOKEN
+# ALLOW_ADVERTISE_STARTD = condor_pool@*/* $(ALLOW_ADVERTISE_STARTD)
+# ALLOW_ADVERTISE_SCHEDD = condor_pool@*/* $(ALLOW_ADVERTISE_SCHEDD)
+# ALLOW_ADVERTISE_MASTER = condor_pool@*/* $(ALLOW_ADVERTISE_MASTER)
+
+# Allow public reads and writes
+ALLOW_READ = *
+ALLOW_WRITE = *
+SEC_READ_AUTHENTICATION = OPTIONAL
+
+ALLOW_ADVERTISE_MASTER = \
+    $(ALLOW_ADVERTISE_MASTER) \
+    $(ALLOW_WRITE_COLLECTOR) \
+    [email protected]
+
+ALLOW_ADVERTISE_STARTD = \
+    $(ALLOW_ADVERTISE_STARTD) \
+    $(ALLOW_WRITE_COLLECTOR) \
+    [email protected]
+
+ALLOW_ADVERTISE_SCHEDD = \
+    $(ALLOW_ADVERTISE_STARTD) \
+    $(ALLOW_WRITE_COLLECTOR) \
+    [email protected]
+EOF
+{{end}}
 # Ohno, not DNS again!
 # 06/18/23 22:49:02 WARNING: Saw slow DNS query, which may impact entire system: getaddrinfo(htcondor-sample-manager-0-0.htc-service.htcondor-operator.svc.cluster.local) took 3.918414 seconds.

controllers/htcondor/templates/execute.sh

@@ -10,6 +10,12 @@ echo "Hello, I am an execute note with $(hostname)"
 # Environment variables specific to submit
 {{template "condor-host" . }}
 
+# Target nodes for dedicated scheduler (allows for parallel universe for MPI)
+# The variable CONDOR_HOST is set in the snippet above
+# https://htcondor.readthedocs.io/en/latest/admin-manual/setting-up-special-environments.html#configuration-examples-for-dedicated-resources
+echo "DedicatedScheduler = \"DedicatedScheduler@${CONDOR_HOST}\"" >> /etc/condor/condor_config.local
+echo "STARTD_ATTRS = $(STARTD_ATTRS), DedicatedScheduler" >> /etc/condor/condor_config.local
+
 # https://github.com/htcondor/htcondor/blob/main/build/docker/services/base/start.sh
 exec bash -x /start.sh
 

controllers/htcondor/templates/manager.sh

@@ -1,15 +1,12 @@
 #!/bin/sh
 
-echo "Hello, I am a server with $(hostname)"
+echo "Hello, I am a config manager node with $(hostname)"
 
 # This script handles shared start logic
 {{template "init" .}}
 
 {{template "config" .}}
 
-export USE_POOL_PASSWORD=yes
-condor_store_cred -p password -f /htcondor_operator/pool_password
-
 # See https://github.com/htcondor/htcondor/blob/main/build/docker/services/base/start.sh
 exec bash -x /start.sh
 

controllers/htcondor/templates/token.sh

@@ -0,0 +1,21 @@
+#!/bin/sh
+
+# Start the server, and keep trying to generate token until it works
+
+export USE_POOL_PASSWORD=yes
+condor_store_cred -p {{.Spec.Config.Password}} -f /root/secrets/pool_password 
+exec bash -x /start.sh &
+
+# Create a token for the execute nodes. 
+# And this will bind to /htcondor_operator and be copied to /root/secrets/token
+retval=1
+while [ ${retval} -ne 0 ]
+  do
+    echo "Waiting for cluster to become ready to generate token...";
+    condor_token_create -authz ADVERTISE_MASTER -authz ADVERTISE_STARTD -authz READ -identity [email protected] -token dockerworker_token
+    retval=$?
+    sleep 2
+  done
+echo "HTCondor properly configured"
+echo "CUT HERE"
+cat /etc/condor/tokens.d/dockerworker_token