cppforlife / checkman Goto Github PK

Cloud Ops just redeployed Concourse, and the number of greys is off the screen, both in the webpage and the smaller popups. On both, it would be good to ensure visibility of the failed builds before the builds that haven't been run yet.

If I use basic auth in the url for checking and it includes a quotation mark, it has to be manually escaped.

Especially since I've moved my notifications to Notification center, seeing the "no icon" icon is a pain.

I'd make a pull request and give you one, but the Icon is the app's personality...

would be pretty sweet

https://github.com/caskroom/homebrew-cask

For an OS X application, the canonical place for this sort of data is in ~/Library/Application Support or ~/Documents, not in ~/Checkman

results in webui showing lost connection...

So there a couple of things that are wrong. The first is that I put a comma on the line, because I thought the config file was Ruby and Shell. My bad, perhaps documentation bad? Since comments are prefixed with hashes anyway, include the shebang on line 1 of the example file so we all know what type of file we're dealing with.

The error is because whatever is processing the config file is firing up an interactive shell, somehow.

$ cat mine
sample: test.check "true", "true"
$

{"result": false, "changing": true}

stty: stdin isn't a terminal

I am seeing a problem with the CircleCI JSON check where it fails to load status for one branch but is able to load the status for another branch. Using the CircleCI status only check works for this branch. I tried running the curl requests directly and saw that the sizes of the results differed dramatically (125KB for working check vs 263KB for the broken one). I don't see any error messages when I run Checkman in the terminal so that I can see its output (all I see is periodic FSChangesNotifier - didNoticeChangeToPath: /Users/username). If the size is the source of the problem, possibly limiting the count of pulled builds would solve the problem (the JSON endpoint supports a limit=10 parameter).

I'm seeing the below error in the checkman debugger window. It appears that the misbehaving character is the long dash thing — '\xE2'

/Users/mopub/.rvm/rubies/ruby-1.9.3-p392/lib/ruby/1.9.1/json/common.rb:148:in encode': "\xE2" on US-ASCII (Encoding::InvalidByteSequenceError) from /Users/mopub/.rvm/rubies/ruby-1.9.3-p392/lib/ruby/1.9.1/json/common.rb:148:ininitialize'
from /Users/mopub/.rvm/rubies/ruby-1.9.3-p392/lib/ruby/1.9.1/json/common.rb:148:in new' from /Users/mopub/.rvm/rubies/ruby-1.9.3-p392/lib/ruby/1.9.1/json/common.rb:148:inparse'
from /Applications/Checkman.app/Contents/Resources/jenkins_build.check:98:in initialize' from /Applications/Checkman.app/Contents/Resources/jenkins_build.check:156:innew'
from /Applications/Checkman.app/Contents/Resources/jenkins_build.check:156:in latest_status' from /Applications/Checkman.app/Contents/Resources/jenkins_build.check:171:in

This causes checkman to blow up on certain checks.

Might just need a line in the ruby app to support UTF-8?

Thanks.

a JSON file in my Checkman directory instead of "defaults write" would be nice...

When trying to view the web UI in Firefox, calls to /check_updates fail with a "400 Bad Request".

Maybe we should use /usr/bin/ruby for all the scripts? The system ruby is more stable and boring.

Currently, we use site checks as a "MVP CI" - e.g. we hit a "canary page" that indicates if a given site is working correctly.

However, we want these to be at a longer interval (e.g. 120 seconds), so they don't spam the server with requests.

The problem is when we add or change a site check name, all existing workstations don't have the local override for the new site name, they default to frequently checking it and spamming it.

This has a "viral" effect, of all workstations spamming as soon as they pull the new Checkfile, until the local interval override is made for the new sites.

Two possible solutions:

1. Make local interval overrides take wildcards/regex, e.g. so all site checks can be given a longer interval, regardless of their name
2. Allow interval to be overridden in the Checkfile itself, so the interval is set as soon as the new Checkfile (and new sites) are checked out.

We have developers both using OSX and linux environment.

What's the best way to port the checkman also on linux environment?

Is there any command line interfacoe for checkman?

Thanks, Shin

The defaults included for checkman have been hammering the Jenkins project's Jenkins instance that I have long since inserted a custom redirect to this page

This traffic is abusive and fills our access logs with garbage requests. I would like to kindly ask that you remove our public Jenkins instance from the defaults for Checkman.

Maybe an About menu item that shows version info? Could also be the place where you put a new version indicator?

We committed once, a build started, and then we pushed another commit. Checkman's information radiator marked the build as red even though it hadn't failed.

@ematpl and I saw this, happy to chat in person if required.

The URL used to check for project errors, http://my-project.airbrake.io/projects/{PROJET-ID}/errors.xml?auth_token={TOKEN}, now returns a 404.

Not sure where you got the original documentation for this, but according to the API overview:

To get the XML for the 30 most recently active errors:
http://your_account.airbrake.io/errors.xml?auth_token=TOKEN

I don't see how to specify a specific project in this URL, but I just skimmed through the docs.

r.

Looks like Xcode iterations have deprecated things that we depend on. Mostly notably, Cedar doesn't seem to be compatible with the new build system.

Putting the build system into a legacy mode helps but Cedar's output is still missing after things run successfully.

Checkman seems to stop polling when using the site script and the response is a 504 (I believe this happens with a 502 too, but didn't have logs to show it). This results in a red build until Checkman is restarted.

Example log output from Alt+clicking the build in question:

cd /Users/pivotal/Dropbox/front-end; /bin/bash "-lc" "PATH=$PATH:/Applications/Checkman.app/Contents/Resources site.check https://console.run.pivotal.io/heartbeat"

{"result":false,"changing":false,"url":"https://example.com/heartbeat","info":[["Url","https://example.com/heartbeat"],["Status","HTTP/1.1 504 Gateway Timeout"],["Content-length:0"]]}

stty: stdin isn't a terminal
stty: stdin isn't a terminal
curl -sIk -A "CheckmanSite (Hostname: $HOSTNAME)" 'https://example.com/heartbeat'
HTTP/1.1 504 Gateway Timeout

Content-length:0

Semaphore has seemingly changed the structure of their project IDs from the style of "3c230d7617c3732dcccccccb1ddc89fc357cd83" to a newer "09e46d07-bbbb-aaaa-a1b0-483212809670" style. This seems to have broken Checkman for new semaphore projects.

should be red and not building:

{
    "id":19198253,
    "repository_id":1377279,
    "number":"103",
    "state":"finished",
    "result":null,
    "started_at":"2014-02-19T17:27:23Z",
    "finished_at":"2014-02-19T17:28:29Z",
    "duration":66,
    "commit":"34d843180f451dc26c690b2b21c4e7243c2f2974",
    "branch":"master",
    "message":"update etcd to 0.3",
    "event_type":"push"
}

We are using the lovely CI in the cloud over at codeship.io. They have a JSON API: https://www.codeship.io/documentation/api-and-webhooks/api/

I tried adding the config setting as described but I still just get custom notifications.

Spotted an oddity with checkman today - it's reporting that one of our builds, redis-release has failed, but when we go to the build server, it's passed.

Looking at what GoCD returns, for our project I see the following:

<Project name="redis-release :: <...>" activity="Sleeping" lastBuildStatus="Success" lastBuildLabel="<...>" lastBuildTime="2015-06-13T12:09:59" webUrl="<...>" />

However, there is another project with the following:

<Project name="cf-redis-release :: <...>" activity="Sleeping" lastBuildStatus="Failure" lastBuildLabel="<...>" lastBuildTime="2015-06-13T12:55:58" webUrl="<...>">

Looking at cctray.check, I guess it's the following code causing the problem:

projects = @feed.xpath('//Project').select do |proj|
  proj.attr('name').downcase.include?(name.downcase)
end

Does that need to be include? or would start_with? or == be OK too?

May also be showing up because cf-redis-release has a more recent build than redis-release.

Using vi to edit a Checkfile while Checkman is running ends up increasing the counter in the system bar, but the UI only shows the updated values. Perhaps the old checks stay around?