scobal / seyren Goto Github PK

This will be useful, oui.

Will be good to see what has happened along with what is happening

Allow us to alert on a check to a specific room on a specific server.

Put as much detail as possibly into email notifications

Just like subscription on the check page but for checks on the checks page

I don't think we're going to need these as strings, lets go for floats

Call the state 'Exception'

Might be nice to have the ability to customize the title and colour of the thresholds. For example instead of just error and warn we could have fyi, wtf and omg.

We should do this every minute or so

Most important:

checks.html --> checks table
home.html --> warn and error tables
check.html --> subscriptions

Less important:

checks.html --> alerts table

If graphite is unavailable we should make sure the UI shows us on every page

After our discussion earlier it might be cool to show people how to do alerting on waves:

timeshifted metrics
holt winters

I suspect this will be useful as a starting point for basic examples too

As a user I want to be able to manually run a check through the ui

We should draw the target along with the warn and error levels on some graphs (maybe last hour, last day, last week?)

We should apply the check to each metric in the response and generate separate alerts.

In case we fire lots of alerts in a short space of time.

Maybe send 10 notifications then an email explaining that no more emails will be sent

How about we add a small graph (similar to the ones on the check page) to the create check form?

This should help make sure a valid graphite target has been entered and also help to set the correct warn and error levels

We could be much more colourful on the home page, the checks page and the check page.

Allow us to choose:

Day (e.g. checkboxes for Sun, Mon, Tue, Wed etc.)
Time (07:00 to 17:00 on)

Find a nice way to hide these forms away by default. Maybe a button "Create check" that opens a form...

Before we release we need to make the following configurable:

1. Graphite server (used server side and client side)
2. SMTP server

We should correctly handle the scenario where the lower a lower value indicates a worse situation

If a check is firing alerts I'd like to be able to add notes to the check.

eg;

check-db-server-is-up - ERROR
MP - We're looking into it
A4J - Bon

Or possibly grey them out.

To enable nice auto-updating and nicer display we should move the check editing from being in place to it's own modal form

Mainly need for check.html but might also be needed for the front page

Currently we notify based on the previous state of the check but we really need to notify based on the previous state of the specific alert for the given target

Currently we can just go to older or newer alerts.

We should do proper paging including page numbers, total results, oldest, newest, etc,.

We currently don't display the result of hitting a button (save, delete). Also - enabling and disabling, paging, etc,.

We should build in a generic feedback (positive and negative) framework into main-controller

We should update the fromState field whenever:

1. We go from OK to WARN.
2. We go from OK to ERROR.
3. We go from WARN to ERROR.
4. We go from ERROR to WARN.
5. We go from WARN to OK.
6. We go from ERROR to OK.

When you add a subscription the subscriptions list should auto-update. It's coded to do this but it's not currently working (consistently)

At the moment the notification service will send a message to each subscriber when an alert is fired.

I'm not sure if we should store every alert when the state is the same but we probably shouldn't send notifications unless the state changes.

Make it easy to disable notifications for a subscription

I suppose we need to show any checks at WARN or ERROR (possibly along with their graphite graphs for sexy times)

@neilprosser any chance you fancy having a crack at this?

My basic idea would be to just run all the checks every 10 seconds (presuming that under ordinary use we wouldn't end up creating a huge amount of alerts). Do you think we need something more?

Might be nice to have the ability to disable alerts for a length of time (for maintenance etc.) and have them automatically turned back on.

It might be useful to put a message like "This check has no subscriptions" in place when there are no subscribers

I think it might be useful to create a concept of groups. As it currently stands we might end up with say 10-20 checks per service. If we added groups of checks we could make the checks more manageable through the ui. Something like:

• live-api
  • live-api-cpu
  • live-api-mem
  • live-api-rqps
• live-solr
  • live-solr-heap

We could even group them by naming convention......

Currently the first field in the form (name) doesn't get focus after clicking Create Check

Bootstrap modal for confirming check deletion and subscription deletion