#mcollective

[mcollective-package-agent] richardc opened pull request #15: Add a yum provider (master...cleanup_pr_3) http://git.io/_jHjRw

[mcollective-package-agent] richardc closed pull request #3: added yum provider (master...master) http://git.io/qermRQ

[mcollective-package-agent] ploubser pushed 3 new commits to master: http://git.io/Ze9itA

mcollective-package-agent/master 4c0ae17 Lars Sjöström: Add a yum provider...

mcollective-package-agent/master 040ab29 Richard Clamp: Add rspec testing to the yumpackage provider...

mcollective-package-agent/master 1b4e808 Pieter Loubser: Merge pull request #15 from richardc/cleanup_pr_3...

[mcollective-package-agent] ploubser closed pull request #15: Add a yum provider (master...cleanup_pr_3) http://git.io/_jHjRw

Hello

is there a way to retrieve the timed out nodes in a usable fashion (json for ex)?

and a second question: is there a way to use mco ping with --dm ?

if you write your own client code you can get the non responsive nodes yeah - stored in an array on the stats thing

mco ping doesnt speak to discovery methods

it tests only the most basic network protocol - the broadcast mode at that. so custom discovery methods cant work

so what's the best way to ping machines taken from a dm?

what are you trying to do? 'mco ping' is a diagnostic tool, its about as usual as icmp ping is in the general case

if u want to do a RPC request and ping that, use rpcutil#ping

I want to get machines from a dm and check if they reply

ok

mco rpc rpcutil ping

so mco rpc rpcutil ping --dm etc.

yeah

thanks

and if you did that in your own code you'd get non responsive reports and all that

that said, it would be useful if the report of unresponsive machines was more readable

machines names tend to touch each other

if they're too long

yeah

Volcane: is there any builtin way (i.e. with the puppet agent) to timeout the whole run (i.e. not just discovery or talking to a server)?

no once things have started on the agent the only way the stop is with the timeout declared in the ddl

Volcane: I'm interested in timing out the client...

like, the whole thing

--timeout ?

`mco puppet runall --timeout X 5` ... and in X seconds, I want the whole process to die if it's not done...

I tried it, and the process ran on happily long past the timeout

yeah so timing out the client just makes it not wait any longer

but like i said above

once something is started on the agent the only thing to stop it is the ddl defined ttl

right, I just don't want it to wait any longer

then -t

Volcane: am I missing something here: http://dpaste.com/3SVM57G

vi un

shouldn't that have exited at 07:53:01 ?

sorry

ah, runall, hmm, it might not really care for that setting :)

yeah.... I was afraid of that...

ok thanks. I'll have a peek at the code...

Volcane, something like https://gist.github.com/raphink/a752dd3d4c51a78... ?

raphink: mc.stats.noresponsefrom

ah, even better :-)

do I need the ping then?

yes

make request, get stats from request on who didnt reply

yup

thanks

[marionette-collective] ploubser pushed 2 new commits to master: http://git.io/HWBUgg

marionette-collective/master 4550d39 Richard Clamp: MCO-411 remove examples of catching RPCErrors in response blocks...

marionette-collective/master 284c0c6 Pieter Loubser: Merge pull request #212 from richardc/MCO-411_docs_for_MCO-264...