Problem: high CPU usage on server
Symptoms:
- You notice high CPU usage on your GHES instance.
- Git or other operations begin to slow down unexpectedly.
Refactor automation
- Every API call and Git operation puts load on the server.
If there are automated tools making frequent calls, consider:
- dialing back their frequency
- running them at off-peak hours
- adding jitter to their start times
- For triggered work, switch from polling to webhooks.
- Redesign your automation to avoid write amplification. For example, if every user push kicks off some kind of automated workflow which itself does a push, you’ve doubled (or more) the impact of each user. The act of writing causes even more writes to be scheduled, hence each write is amplified.
Use caution with, or don’t use, shallow clones
- Shallow clones with a depth greater than 1 are computationally expensive.
Switch your workflow to use shallow clones with
--depth=1. - Fetching from shallow clones is also computationally expensive. We recommend using partial clone if you need to query the commit history.
- Unshallowing a shallow clone is also computationally expensive. We recommend using a full clone or a partial clone in lieu of unshallowing.