Collection Health

The DBA Dash service reports every collection run to the repository. Collection Health shows whether your collections are working, how much they cost, and how long their data waits before it’s imported.

Where to find it#

• Collection Health tab under the Checks node of the tree.
• Collection Health column on the Summary tab. Click a cell to open the Collection Health tab for that instance.

Collection Health and Collection Dates#

DBA Dash checks collections in two ways, and both are shown on the Summary tab:

Collection Health can only report on runs that happen. If a collection stops running altogether - the service is stopped, an instance has been removed from the service config, or a schedule is broken - there are no failed runs to report. Snapshot Age catches this. It highlights a collection whose last run is older than its threshold.

On the Collection Dates tab under the Checks node, a collection with no threshold set uses one based on its schedule: warning at twice the longest interval between runs plus 3 minutes, and critical at 4 times the interval plus 6 minutes. You can set your own thresholds for all instances or for one instance. See Collection Thresholds.

Use both: Snapshot Age tells you a collection has stopped, and Collection Health tells you a collection is running but not working.

Run outcomes#

Each collection run is recorded with one of these outcomes:

Some collections have steps that run after them, such as collecting query text and plans for running queries and query stats. These steps are recorded as runs of their own.

The report#

The report shows a row per collection per instance, or a row per collection across all the selected instances. It can include or leave out the steps that follow some collections. Columns include:

• Success rate, failures in a row, retries and partial runs
• Status, and the reason for it
• Collect and import duration
• CPU and logical reads, on the monitored instance and in the repository
• Import lag

Hover over a column heading to see which period it covers. Failures in a row show the current state, whatever period is selected. A weekly collection that failed at the weekend still shows as failing over the last hour.

Charts show runs by outcome, cost and import lag over time. Where each run is kept (see Cost), the charts go down to 5 minute buckets.

The counts link to the errors behind them. Errors, Failed and Import Failed leave out retries, and Retried includes them. Last Failure links to the errors for that failure, whatever period is selected.

Collection Performance#

Click a collection to open a Collection Performance report, which shows each run (or each hour) for that collection over time.

Status and thresholds#

A collection’s status is OK, Warning or Critical. Two things can raise it.

Failures in a row#

Failures in a row only raise a status when they reach both a number of failed runs and a time since the last success:

A collection that runs every minute has to keep failing for 10 minutes before it shows. A collection that runs daily shows on its first failure. If a collection has never succeeded, only the number of failed runs is used.

Failure rate#

The failure rate is used when there are at least 5 runs:

A collection that has recovered is a warning at most. On the Summary tab, the failure rate covers the time since errors were last acknowledged, up to 24 hours.

Changing the thresholds#

Click the Thresholds column to change the thresholds. They can be set for:

• a collection on one instance
• every collection on an instance
• a collection on every instance
• every collection on every instance

The most specific setting wins, one threshold at a time. Anything not set uses the defaults above. Select Don’t monitor to stop a collection affecting the status.

Paused instances and disabled schedules show as N/A.

Acknowledging errors#

Acknowledge Errors on the DBA Dash ErrorLog toolbar clears failures in a row until the next failure.

Older services#

Instances whose DBA Dash service is older than 4.25 don’t report collection runs. For those, the Summary tab counts collection and import errors in the last 24 hours, as before.

Cost#

Turn on Log Internal Performance Counters in the service config tool to record cost. With it on:

• Each run is kept, as well as the hourly rollup.
• CPU and logical reads are recorded for each collection on the monitored instance.
• CPU, logical reads and duration are recorded for importing each collection into the repository.

Without it, only the hourly rollup is kept, CPU and logical reads aren’t recorded, and the charts use hourly averages.

Import lag#

Import lag is the time from the service handing a batch of collected data to its destination to the import starting.

• When the service writes straight to the repository database, import lag is near zero.
• When the destination is an S3 bucket or a folder, import lag is the backlog waiting to be imported.

Metrics and alerts#

Collection Health data is also available as DBA Dash counters:

These can be charted and used in alert rules like any other counter. On the Metrics tab they’re hidden by default. Click Collection Metrics to show them. Collection Duration (ms) keeps the same CounterID it had before 4.25, so existing alert rules on it carry on working.

DBA Dash ErrorLog#

The DBA Dash ErrorLog tab under the Checks node lists the errors behind Collection Health. It opens on a row per instance, or a row per source when one instance is selected. The counts are split two ways:

• Errors / Warnings / Retries - what kind of error
• Collect / WMI / Import / Other - where it happened

Click an instance, a source or a count to narrow the report, and Back to go up again.

WMI suggestions#

When 2 or more collections on an instance fail with WMI errors, WMI is probably not working on that instance at all. Collection Health and the ErrorLog show a card suggesting that you fix WMI access, or select No WMI for the instance on the Source tab of the service config tool. See WMI.

Data retention#

See Data retention to change these.