What are performance archives and how are they triggered?
- Performance content is collected and stored (archived) on each system
- Metrics like node and volume performance statistics are logged
- Data is kept on the root volume for a rolling 28 days by default
- Data is constantly logged, so there is no starting or stopping advanced logging
- There is minimal overhead to keep latency down
- Using the AutoSupport(tm) infrastructure, a Performance Archive can be created and uploaded to NetApp
- Performance archives do not replace performance monitoring or help with sizing as they are intended to troubleshoot ONTAP performance issues
How to upload a Performance Archive?
- Confirm the AutoSupport transport is HTTP or HTTPS
Cluster::> autosupport show -fields transport
- Performance archives are typically 50 MB or bigger when compressed so ONTAP is hard coded to not transmit if transport is SMTP
- To switch to HTTPS from SMTP which does allow larger file sizes to be trasferred, the command that can be used is:
system node autosupport modify -node * -transport https
- If the transport is SMTP and the other methods do not work or are not allowed in the environment, use the manual method with 7zip files or use OneCollect to gather the data
- Run the
system node autosupport invoke-performance-archivecommand
Cluster::> system node autosupport invoke-performance-archive -start-date "11/31/2022 06:00:00" -duration 4h -case-number 2008123456 -node *
- Here are the following options for the command:
-start-date <"MM/DD/YYYY HH:MM:SS">
-duration <[ h][ m][ s]> (or -end-date)
- It is important to note that the maximum duration for any single collection is 6 hours and the recommended sample period is 4 hours
- Anything less than 2 hours will miss some hourly counters
-start-datemust be at least the specified
-durationhours in the past, otherwise complete time-frame will not be collected.
- No error message will be displayed indicating that complete
-durationwas not collected
- It is best to start on the top of the hour and end on the top of the hour to capture the hourly counters
- The whole cluster must be triggered since workload statistics are measured across the cluster
- Ideally getting the hour before and after the performance event is recommended
- Date format is recommended to be quotes
- To view the progress of the Performance Archive upload you can use the following commands:
::> autosupport history show-upload-details -node <node name> -seq-num <seq-num>
What if performance archive doesn't work?
- Troubleshoot Autosupports using the ONTAP AutoSupport Resolution Guide
- Verify if Performance Archive is enabled and the date of the archive you're requesting is less than retention period(28 days by default)
::> set diagnostic
::*> statistics archive config show
Is the Performance Archive Enabled?: true
if false, enable it by running the following command
::*> statistics archive config modify -is-enabled true
Note: EMS events will also report the following message when the Performance Archive is disabled
::*> event log show -event perf*
Time Node Severity Event
------------------- ---------------- ------------- ---------------------------
INFORMATIONAL perf.ccma.off: Performance archiver is not enabled.
Why would perfstat be needed?
- Support will request a perfstat
- On ONTAP releases up to 9.5, some data may be missing and Support will ask for a perfsat
- Only gather if requested by Support
- Sometimes the sales team may request to gather performance data for sizing
How can a storage administrator look at the content within a performance archive?
- Active IQ Unified Manager utilize the underlying data set that is stored within a Performance Archive
- Other tools such as Harvest and Cloud Insights also pull the same data and may be used
- The data is a binary file that cannot be viewed manually and one of the above tools must be used