Changes between Version 45 and Version 46 of NeoContainers

Timestamp:

Jan 27, 2016 11:09:51 AM (9 years ago)

Author:

Geoff Lawler

Comment:

--

NeoContainers

@@ -695 +695 @@
 source tb_compat.tcl
 
-tb-make-soft-vtype container0 {pc2133 MicroCloud dl380g3}
+tb-make-soft-vtype container0 {pc2133 MicroCloud}
 
 set base1 [$ns node]
@@ -717 +717 @@
 Create a new experiment with the NS file above. *Do not* swap in the experiment. 
 
-We run our own configuration database on the {{{chef}}} node. This is not required, but keeps our configuration separate from the "official" configuration database.
+We run the configuration server on the {{{chef}}} node by hand in order to confirm things are working. Do the following in a separate window so we can see the log output and confirm things are working. If there is a {{{config_server}}} already running, kill it first.
 
 {{{
@@ -724 +724 @@
 chef: > cd src/config_server
 chef: > git pull ~chef-user/config_server   # grab the newest code from chef-user.
-chef: > ./runserver.py -l debug -p 5323 -d $(pwd)/config.db    # full path to DB and unused port.
+chef: > ./runserver.py -l debug -d $(pwd)/config.db    # full path to DB
 }}}
 
@@ -730 +730 @@
 
 {{{
-#!json
+#!python
 [
         {
@@ -759 +759 @@
 Note that the Win7 nodes are in the same subnet at the physical nodes.
 
-Load the configuration into our configuration server running on {{{chef}}}. This must be done *before* swapping in the experiment as the configuration server must request control network addresses from DETER. 
-
-{{{
-#!sh
-users: > /share/config_server/bin/initialize_containers.py -p deter -e neo-win7 -f neo-win7.nodes.json -P 5323 -l debug
-}}}
-
-Note that {{{neo-win7.nodes.json}}} is the JSON file above and we use the port of our own config_server.
+Load the Win7 nodes' configuration into our configuration server running on {{{chef}}}. This must be done *before* swapping in the experiment as the configuration server must request control network addresses from DETER. 
+
+{{{
+#!sh
+users: > /share/config_server/bin/initialize_containers.py -p deter -e neo-win7 -f neo-win7.nodes.json  -l debug
+}}}
+
+Note that {{{neo-win7.nodes.json}}} is the JSON file above and we use the port of our own config_server. You should see log activity in your {{{config_server}}} window.
 
 Output should not contain {{{[ERROR]}}} or {{{[WARN]}}} messages. Here's an example of a good run:
@@ -775 +775 @@
 01-27 09:29:11 __main__     DEBUG    using base url: /exp/deter/neo-win7/insert
 [OK   ] : Initializing containers from neo-win7.nodes.json for use in experiment deter/neo-win7
-01-27 09:29:11 __main__     DEBUG    sending url: http://chef:5323/exp/deter/neo-win7/insert/containers
-[OK   ] : upload_node_file: Connecting to server chef:5323
+01-27 09:29:11 __main__     DEBUG    sending url: http://chef:5320/exp/deter/neo-win7/insert/containers
+[OK   ] : upload_node_file: Connecting to server chef:5320
 [OK   ] : upload_node_file: Posting node data ... url=/exp/deter/neo-win7/insert/containers
 [OK   ] : Upload_node_file: Getting Response
@@ -783 +783 @@
 }}}
 
+You should see the win7 nodes from your {{{nodes.json}}} file added. (This is the "Nodes added: billg, ballmer" above.)
+
 Now swap in the experiment:
 
@@ -790 +792 @@
 }}}
 
-Once swapped in, we log into a base node and kick off the chef processes to create and start the Win7 node.
+Once swapped in, we give the swap-in-specific details to the configuration database (groups, users, mounts, etc). This only needs to be done once per experiment swapin. We do this so the containers know the experiment/swapin specific information. We cannot do this before swapin as the information does not exist then!
+
+{{{
+#!sh
+users: > ssh base1
+base1: >  /share/config_server/bin/initialize_configuration_database.py -p deter -e neo-win7 -l debug   # note same port and chef server.  
+}}}
+
+You should see log messages in your {{{config_server}}} window, user ids, mounts, group information for the experiment will scroll by.
+
+Successful output will contain one {{{[WARN]}}} message. We get this warning because the script did not find a {{{proj/deter/exp/neo-win7/containers}}} directory. This directory is created by the old containers system. Because we're not using that system in this example, this warning is OK. Here is the successful output:
+
+{{{
+#!sh
+OK   ] : Initializing configuration database from experiment deter/neo-win7 data.
+[OK   ] : Posting mounts data...
+01-27 10:18:51 __main__     DEBUG    posting url: /exp/deter/neo-win7/insert/mounts
+[OK   ] : Connecting to server
+[OK   ] : Good response from server.
+[OK   ] : Success loading mount information.
+[OK   ] : Posting accounts data...
+01-27 10:18:51 __main__     DEBUG    posting url: /exp/deter/neo-win7/insert/accounts
+[OK   ] : Connecting to server
+[OK   ] : Good response from server.
+[OK   ] : Success loading account information.
+[OK   ] : Posting physical node names data...
+01-27 10:18:51 __main__     DEBUG    posting url: /exp/deter/neo-win7/insert/pnode_names
+[OK   ] : Connecting to server
+[OK   ] : Good response from server.
+[OK   ] : Success loading physical node names.
+[WARN ] : No containers information found at /proj/deter/exp/neo-win7/containers. Not loading containers information.
+[OK   ] : All is well. Configuration database now contains information about experiment deter/neo-win7.
+}}}
+
+Now we confirm that chef is ready to run, by confirming that the cookbooks are loaded into the running chef instance. (This does not need to be done if you know chef is configured properly. This is here for completeness.)
+
+{{{
+#!sh
+users: > ssh chef
+chef: > sudo su - chef-user
+chef: > cd ~/chef-repo
+chef: > knife cookbook list
+}}}
+
+This will show about 5 to 7 cookbooks loaded. If not, load them. (Note you can reload them if you want to be sure all is well.)
+
+{{{
+#!sh
+users: > ssh chef
+chef: > sudo su - chef-user
+chef: > cd ~/chef-repo
+chef: > knife cookbook upload --all
+}}}
+
+
+Now everything is set to start chef on the base node. On each base node, run the {{{bootstrap_node.sh}}} script to bootstrap the base node into the chef system. Once bootstrapped, we can run {{{chef-client}}} to configure the base node. 
+
+{{{
+#!sh
+users: > ssh base1
+base1: > /share/config_server/bin/bootstrap_node.sh -s chef -p 5323 -r pnode
+}}}
+
+When running, you should see logging in the {{{config_server}}} window. 
+
+Now start chef-client on the base node.
+
+{{{
+#!sh
+users: > ssh base1
+base1: > sudo chef-client
+}}}
+
+
+There will be a bunch of output. You can see various chef cookbooks loaded and executed. The output will pause for a long time on {{{execute[vagrant up --parallel] action run}}}. This statement is the virtual machines booting on the base node. This will take a *long* time, 10 minutes or more. This step is starting the win7 virtual nodes *and* having the win7 nodes also chef-bootstrap and chef-configure themselves. Unfortunately there is no easily observable log output for these events. Once {{{chef-client}}} completes, the Win7 nodes should be up and accessible in the experiment. 
+