slaprunner: fix resiliency test, add development-mode (bool) parameter for...

slaprunner: fix resiliency test, add development-mode (bool) parameter for easy build webrunner in webrunner

software/slaprunner/common.cfg

@@ -47,14 +47,14 @@ parts =
 recipe = slapos.recipe.template
 url = ${:_profile_base_location_}/instance.cfg
 output = ${buildout:directory}/template.cfg
-md5sum = bb7e0bf9959c4437ff1e23e645315ccf
+md5sum = 37fe35675039e29c20838aedb8005fd0
 mode = 0644
 
 [template-runner]
 recipe = slapos.recipe.template
 url = ${:_profile_base_location_}/instance-runner.cfg
 output = ${buildout:directory}/template-runner.cfg.in
-md5sum = 5fbdf6f9996d6cb948ba042e9dd6e43e
+md5sum = e0661b2532ecd36a56f18b7390d25599
 mode = 0644
 
 [template-runner-import-script]
@@ -101,7 +101,7 @@ recipe = hexagonit.recipe.download
 ignore-existing = true
 url = ${:_profile_base_location_}/instance-resilient-test.cfg.jinja2
 download-only = true
-md5sum = fb8c45e5c35548331fb06c4633ec592a
+md5sum = 51248bb1bbf54ad11d0674ba40be18b4
 filename = instance-resilient-test.cfg.jinja2
 mode = 0644
 

software/slaprunner/instance-resilient-test.cfg.jinja2

@@ -42,8 +42,8 @@ mode = 755
 # Used to manually run the resilient test if we don't have a running testnode.
 recipe = slapos.cookbook:wrapper
 test-suite-title = slaprunner
-command-line = {{ bin_directory }}/runStandaloneResiliencyTestSuite --test-suite-title=${:test-suite-title} ${deploy-scalability-test:test-parameters}
-wrapper-path = ${directory:bin}/runStandaloneResiliencyTestSuite
+command-line = {{ bin_directory }}/runStandaloneResiliencyTest --test-suite-title=${:test-suite-title} ${deploy-scalability-test:test-parameters}
+wrapper-path = ${directory:bin}/runStandaloneResiliencyTest
 
 [request-resilient-instance]
 <= slap-connection
@@ -57,10 +57,11 @@ config-{{ key }} = {{ dumps(value) }}
 config-resiliency-backup-periodicity = */10 * * * *
 config-resilient-clone-number = 1
 config-ignore-known-hosts-file = true
+config-development-mode = {{ slapparameter_dict.get('development-mode', '') }}
 # XXX hardcoded
 #config-frontend-domain = google.com
 # XXX Hack to deploy Root Instance on the same computer as the type-test Instance
 sla-computer_guid = ${slap-connection:computer-id}
-return = backend_url
+return = backend-url
 
 [slap-parameter]

software/slaprunner/instance-runner.cfg

+{% set runner_port = 50005 -%}
+{% set proxy_port = 50000 -%}
+{% set dropbear_port = 22222 -%}
+{% set supervisord_port = 39986 -%}
+{% set runner_http_port = 9686 -%}
+{% set monitor_http_port = 8386 -%}
+{% set runner_software_type_list = ['runner-export', 'runner', 'default'] -%}
+
+{% if slapparameter_dict.get('development-mode', '') == 'true' -%}
+{%   if slap_software_type in runner_software_type_list -%}
+{%     set runner_port = 50025 -%}
+{%     set proxy_port = 50020 -%}
+{%     set dropbear_port = 22242 -%}
+{%     set supervisord_port = 39996 -%}
+{%     set runner_http_port = 9696 -%}
+{%     set monitor_http_port = 8396 -%}
+{%   elif slap_software_type == 'runner-import' -%}
+{%     set runner_port = 50026 -%}
+{%     set proxy_port = 50021 -%}
+{%     set dropbear_port = 22243 -%}
+{%     set supervisord_port = 39997 -%}
+{%     set runner_http_port = 9697 -%}
+{%     set monitor_http_port = 8397 -%}
+{%   endif -%}
+{% endif -%}
+
 [buildout]
 parts =
   nginx_conf
@@ -141,8 +167,8 @@ private_key = $${sshkeys-dropbear-runner:private-key}
 ipv4 = $${slap-network-information:local-ipv4}
 ipv6 = $${slap-network-information:global-ipv6}
 instance_root = $${runnerdirectory:instance-root}
-proxy_port = 50000
-runner_port = 50005
+proxy_port = {{ proxy_port }}
+runner_port = {{ runner_port }}
 partition-amount = $${slap-parameter:instance-amount}
 wrapper = $${directory:services}/slaprunner
 debug = $${slap-parameter:debug}
@@ -212,7 +238,7 @@ keygen-binary = ${dropbear:location}/bin/dropbearkey
 [dropbear-runner-server]
 recipe = slapos.cookbook:dropbear
 host = $${slap-network-information:global-ipv6}
-port = 22222
+port = {{ dropbear_port }}
 home = $${buildout:directory}
 wrapper = $${directory:bin}/runner_sshd
 shell = ${bash:location}/bin/bash
@@ -542,12 +568,12 @@ auto-deploy-instance = true
 autorun = false
 slaprunner-httpd-port = $${:monitor-port}
 # XXX - for backward compatibility, monitor-port was for slaprunner httpd server
-monitor-port = 9686
+monitor-port = {{ runner_http_port }}
 instance-name =
 monitor-cors-domains = 
 monitor-interface-url = 
 # XXX - define a new port for monitor here and use monitor-port for backward compatibility
-monitor-httpd-port = 8386
+monitor-httpd-port = {{ monitor_http_port }}
 
 [monitor-parameters]
 port = $${slap-parameter:monitor-port}
@@ -671,7 +697,7 @@ numprocs = 1
 path = $${shell:path}
 pidfile = $${directory:run}/supervisord.pid
 server = $${slaprunner:ipv4}:$${:port}
-port = 39986
+port = {{ supervisord_port }}
 slapgrid-cp = slapgrid-cp
 slapgrid-cp-command = $${slaprunner:slapos} node instance --all --cfg $${:slapos-cfg} --pidfile $${:slapgrid-cp-pid} --verbose --logfile $${:slapgrid-cp-log}
 slapgrid-cp-log = $${runnerdirectory:home}/instance.log

software/slaprunner/instance.cfg

@@ -26,6 +26,7 @@ context = key buildout buildout:bin-directory
           key develop_eggs_directory buildout:develop-eggs-directory
           key eggs_directory buildout:eggs-directory
           key slapparameter_dict slap-configuration:configuration
+          key slap_software_type slap-configuration:slap-software-type
           raw curl_executable_location ${curl:location}/bin/curl
           raw dash_executable_location ${dash:location}/bin/dash
           raw dcron_executable_location ${dcron:location}/sbin/crond

stack/resilient/buildout.cfg

@@ -69,7 +69,7 @@ mode = 0644
 [template-replicated]
 recipe = slapos.recipe.build:download
 url = ${:_profile_base_location_}/template-replicated.cfg.in
-md5sum = 7a6234465ae845cb262d4f94c158764e
+md5sum = e66b3f2316ac58348c35541b939b5656
 mode = 0644
 destination = ${buildout:directory}/template-replicated.cfg.in
 

stack/resilient/template-replicated.cfg.in

@@ -90,6 +90,10 @@ config-number = {{id}}
 config-name = {{namebase}}{{id}}
 config-authorized-key = ${request-pbs-{{namebase}}-{{id}}:connection-ssh-key}
 config-on-notification = ${request-pbs-{{namebase}}-{{id}}:connection-feeds-url}${:pbs-notification-id}
+# set development mode
+{% if slapparameter_dict.has_key('development-mode') -%}
+config-development-mode = {{ slapparameter_dict.get('development-mode') }}
+{% endif -%}
 {% for key, value in monitor_dict.iteritems() -%}
 config-{{ key }} = {{ value }}
 {% endfor -%}

[Rafael Monnerat]

I don't like introduce special parameters like this w/o actual meaning for a production instance. We often take this decision, to avoid resolve dynamicaly the conflicts.

I prefer all ports work are dynamic selected from a range, like the one @Nicolas already implemented on his branch for resilience. Please rework this patch to replace this development mode by dynamic port selection.

I don't like use software types and parameters to differ real and test environment. Let's solve the problems with code, it prevent us to document yet another parameters.

[Nicolas Wavrant]

Hello,

Indeed I've been working on a new buildout recipe to select a port from a range, only if it's free. The recipe loops from the lowest port number of the range to the highest one, which give always the same result for the same deployment case (for exemple, for a webrunner deployed on a bare slapos node, the SSHD service will always run on port 22222) : free_port buildout recipe.

To prevent race condition in the case of developping the webrunner inside a webrunner (iow, runner-exporter and runner-importer share the same IP), I use different port ranges for any service running both in exporter and importer. For exemple, in the resilient stack : pbsready-export.cfg.in and pbsready-import.cfg.in. (It seems it works well without always defining different port ranges for different instances, but I should apply this policy everywhere to prevent bad surprises later).

As we are using potentially a wider range of ports than before, review should focus on "could it possibly use a port from a standard service, as zope server, bdd server, ...".

The advantages are :

• anyone can deploy a webrunner and develop it, without reading document/readme first (because nobody does). Single-click-deployment.
• the deployment recipe is more dynamic : I can deploy a webrunner into a webrunner into a webrunner. Useful if I need to fix slapos.core into my 2 webrunner, to be sure that my new version of webrunner (level 3) deploys correctly.
• even if there still is a conflict when starting the service, at least we had an opportunity to fix it : there is still an improvement for defining a constant as a port (in that case neither nothing can be done). And eventually the sysadmin can change the value in the .installed.cfg file (when nothing could be done before).

The disavantages so far:

• To prevent the port from being recalculated during a "slapos node instance", I read the .intalled.cfg file to see if the recipe was already processed. There should be a cleaner way (but I don't want to save it in yet another file, in a more or less random path).

Clearly we are going in complete different directions, and we should discuss about which one should be picked before any of us do more work on our respective branch.

[Alain Takoudjou]

I think the Nicolas solution is more generic than having a parameter for development mode. I new he had a problem to run webrunner inside webrunner like me but I didn't knew he was working on a solution for it. I did this commit quickly to simplify the development of webrunner recipe because it was always hard to have everything running without port conflict. I will just revert my commit and let Nicolas merge his one changes as he has something already working.

[Rafael Monnerat]

@alain.takoudjou @Nicolas , I think Nicolas would be interested on those changes and as well Alain should be interested on what Nicolas is doing on resilience to develop webrunner inside a webrunner (and to run tests). Please make sure you don't overlap each other and generate design conflicts.

@alain.takoudjou please keep one eye on:

https://lab.nexedi.cn/Nicolas/slapos.toolbox https://lab.nexedi.cn/Nicolas/slapos