2023-05-18: Increased latency for code suggestions
Customer Impact
Current Status
We fixed the latency issue by increasing the replica count for triton
from 1 to 4 since we where CPU saturated.
We see the amount of requests to be the same but the latency is back to a manageable level.
Corrective Action
- We continue to work on the corrective action needs from the previous incident #14451 (closed)
- We look into further actioning on triton optimisation through model analyzer or GPU optimisation through inference . (gitlab-org/modelops/applied-ml/code-suggestions/ai-assist#104 (closed))
- We continue to monitor cpu -gpu and opt for load-balancing.(gitlab-org/modelops/applied-ml/code-suggestions/ai-assist#105 (closed))
- We write infrastructure as code including terraform scripts , kubectl commands that automated de-bugging.(gitlab-org/modelops/applied-ml/code-suggestions/ai-assist#108 (moved))
- Add more instrumentation and timing information to model gateway logs (time spent in gitlab auth, time spent in triton). (gitlab-org/modelops/applied-ml/code-suggestions/ai-assist#107 (closed))
- This is the first , we are prepared to learn and act upon with a sense of urgency and plan for next.
📚 References and helpful links
Recent Events (available internally only):
- Feature Flag Log - Chatops to toggle Feature Flags Documentation
- Infrastructure Configurations
- GCP Events (e.g. host failure)
Deployment Guidance
- Deployments Log | Gitlab.com Latest Updates
- Reach out to Release Managers for S1/S2 incidents to discuss Rollbacks, Hot Patching or speeding up deployments. | Rollback Runbook | Hot Patch Runbook
Use the following links to create related issues to this incident if additional work needs to be completed after it is resolved:
- Corrective action ❙ Infradev
- Incident Review ❙ Infra investigation followup
- Confidential Support contact ❙ QA investigation
Note: In some cases we need to redact information from public view. We only do this in a limited number of documented cases. This might include the summary, timeline or any other bits of information, laid out in out handbook page. Any of this confidential data will be in a linked issue, only visible internally. By default, all information we can share, will be public, in accordance to our transparency value.