hubspot / singularity Goto Github PK

https://host/singularity/tasks/cleaning

you can get a task to show up there by hitting this endpoint:

http post http://host/singularity/v1/requests/request/{requestid}/bounce

it should list all registered webhooks, which can retrieved via /singularity/v1/webhooks

just shows nothing right now.

If the singularity API returns a error like 502, there's no indication that loading the data failed when navigating to a /tail page

in task list, it's very easy to move your cursor up and have the popover stay around

POST

/requests/request/{requestId}/run

@post
@path("/request/{requestId}/unpause")

@b-ash

Searching https://tools.hubteamqa.com/singularity/v1/history/tasks/search?requestIdLike=w currently returns 9.8Mb of data.

(Add some info and assign back to me)

I have a pair of ginormous Thunderbolt displays but I can't cover them in Matrix-style streams of logs because the area is fixed size 😿 plus, no one likes nested scroll areas.

when looking at the active requests page, we should separate scheduled requests from the others in separate tables / sections of the page.

scheduled requests, for instance, should display the schedule and not the # of instances.

we can iterate from there.

We have an audit log of all changes made to a request.

This might include if someone paused it, bounced it, updated it, created it, deleted it (the audit log is still valid for deleted requests which can be found through search.)

This should (probably) be displayed on the request page:

https://{hostname}/singularity/request/{requestId}

This history comes back from the endpoint:

/history/request/{requestId}/requests

It is a list of RequestHistory objects:

https://github.com/HubSpot/Singularity/blob/master/SingularityBase/src/main/java/com/hubspot/singularity/SingularityRequestHistory.java

For the paginated historical tasks table, sorting is done properly from page to page, but bottom to top on a given page instead of top to bottom. (Example page.)

I'll illustrate with a diagram:

Page 1

Page 2

And so forth.

The correct behavior would instead be:

Page 1

Page 2

Let me know if you need help figuring out how to tell.

Should be disabled if 0 historical tasks?

What's the license of this Mesos framework? (I see no LICENSE file or license information in the pom.xml)

Remove should be called "Kill Task" in the tasks listing.

It should say long running process will be started again instantly, scheduled tasks will behave as if the schedule task failed and may be rescheduled to run in the future depending on retry schedule.

There should also be a Kill Task button on the task overview page.

the scheduled tasks tab has it:

but the scheduled tasks table in a Request does not:

this should be included in dashboard

I have a job with a cron schedule * * * * * which should run every minute. In singularity this is displayed as 0 * * * * ?.

It also appears the same in the JSON file for the request. Is this just a difference in how cron schedules are encoded in mesos or is it actually a bug?

looks like it's ordering on the string literal:

the file view will page when scrolling up, and tail when scrolling down, but it still needs to periodically tail when at the end of a file

1. No obvious way to go back in the tree (or location in the tree)

2. No obvious message that a particular directory is empty.

DELETE

/tasks/task/{taskId}

eg,

/tasks/task/test-rack-1384834126092-788-hodge-us_east_1b

DELETE:

/requests/request/{requestId}

@adamschwartz

Unlike Right Click → Open Link in New Tab.

Try it on the Files links on this page.

@adamschwartz

needs to call delete /requests/request/{request_id}/paused

http://heliograph.iad01.hubspot-networks.net:7005/singularity/#request/ApiManagerWeb:3605:web:1384527552228

1. Should show active, scheduled, and historical tasks.

• historical tasks can be fetched here:
  http://heliograph.iad01.hubspot-networks.net:7005/singularity/v1/history/request/ApiManagerWeb:3605:web:1384527552228/tasks
• note that the history now shows the last task status and the time it happened at (this can tell you how long a task ran for, if it's not currently running, for example.)
• for now, active and scheduled can be parsed out of the active/scheduled endpoints.
• each task in these sections should have a column for taskId, createdat, rackId, instanceNo
• history may need to be pageable

2. Should show request history (this is updates / when it was created OR deleted etc.):
   http://heliograph.iad01.hubspot-networks.net:7005/singularity/v1/history/request/ApiManagerWeb:3605:web:1384527552228/requests

need this visible in order to correlate services in Rodan easier

It should also warn the following:

"Removing this request will kill all active and scheduled tasks and tasks for it will not run again unless it is reposted to Singularity"

I wonder if this should return something else, because it can currently return multiple results with the same request id. If the point of this search is to find requests, maybe it should only return distinct requests, or return only updates which are creations, or it should do some sort of collapsing on the backend to return a more complicated object (ie, requestId + lastStatus + lastTime + creationTime, etc.)

The list of files shows that the file isn't empty, and the /tail page gets an API response with data, but nothing is displayed.

It's usually there after refreshing the page.

This page would allow you to search for request history or task history using these endpoints:

/history/tasks/search?requestNameLike=

/history/requests/search?requestNameLike=

The page just scrolls forever when it gets a 500 instead of saying something like:

"The files for this task could not be loaded. The files may no longer exist on the designated slave and may no longer be retrieved."

add:

POST
/webhooks/
an array of strings e.g, [ "http://www.example.com" ]

DELETE
/webhooks/{hook}

eg,.

/webhooks/http%3A%2F%2Fwww.example.com (url encoded)

Each task update should be an entry in a table.

http://heliograph.iad01.hubspot-networks.net:7005/singularity/#task/ApiManagerWeb:3605:web:1384527552228-1384527575683-1-useast1a

would call:

http://heliograph.iad01.hubspot-networks.net:7005/singularity/v1/history/task/ApiManagerWeb:3605:web:1384527552228-1384527575683-1-useast1a

will now be null when it has never been seen. the UI should respond appropriately.

For dead slaves or racks, we should be able to easily remove them from the UI via a remove link:

http delete https://host/singularity/v1/slaves/slave/{slaveId}/dead

{slaveId} looks like 201311111611-3792629514-5050-11268-18

It should be obvious when are you looking at a task vs request as well.

Just like active tasks.

it'd be useful if if we had a view that listed all tasks running on a specific slave (perhaps linked to from the Slaves page)