Forum Discussion

Bastian_Jeske's avatar
7 years ago

Make Instance Groups searchable/filterable

Hello,

We'd like to request some more usage for instance groups. Right now, it's just not very useful to group instances on a datasource. We have shared devices with datasources belonging to different teams and we have to create dashboards and alarm rules regarding those. Right now, we have to use the wildcard filter in a "creative" way to have shared device alerts and dashboards from different teams configured. It would be really helpful if the instance-group name could be used in Filters.

Use-Case:
  * To configure alert rules for shared devices for different teams, we can group all datasource-instances in instance groups named "teamname" and then filter on "teamname", this works even when we use "*" for device/devicegroup, as long as instancegroup "teamname" is persistent over multiple shared devices.
  * To have dashboards for shared devices on a per-team base, we can filter for the teamname when creating those dashboards. This also works with "*" as device/devicegroup query, so instances on new devices will be added automatically.

Regards,
Bastian

2 Replies

Replies have been turned off for this discussion
  • I agree with this.  Being able to filter alerts on instance groups would be particularly valuable to us, because we have some instances grouped by Production/Non-Production, and the ability to separate these two groups into separate emails (Production would generate a page through PagerDuty, for example, while Non-Production would just send an email notification).

  • I just spent a lot of time beating my head on this, and found the problem is that instance groups are not really groups, they are just stored as labels on instances.  As such, they themselves have no place to store additional information, like device groups can.  In addition to the above desirable features, it should be possible to set alert expressions on instance groups that are persistent (like for device groups).  I ended up dealing with the latter issue via the API, so now I can do this:

    ./lm-set-alertexpr --company mycompany --devicegroup "All Exchange Servers" --datasource "WinVolumeUsage-" --datapoint "PercentUsed" --instancegroup "MailboxDB" --instancegroupop '!:' -v -n --alertexpr "> 90 95 99"

    So at least some issues can be resolved via the API, but alert routing via instance group is definitely needed!  Anywhere an instance can be used, an instancegroup should be also possible, definitely.

    Regards,
    Mark