Cloud Providers
Impact statement: Beginning as early as 11 Aug 2023, you have been identified as a customer experiencing timeouts and high server load for smaller size caches (C0/C1/C2).
Current status: Investigation revealed the cause to be a change in behavior of one of the Azure security monitoring services agent used by Azure Cache for Redis. Monitoring Agent subscribes to the event log and has scheduled backoff for resetting subscription in case no events are received. In some cases scheduled backoff is not working as expected and can increase the frequency of subscription resetting which can significantly affect CPU usage for smaller size caches. Currently, we are in progress of rolling out of the hotfix to the impacted regions which is 80% completed. Initially we estimated this to complete by 13 Oct 2023, however, progress shows we are expected to complete by 11 Oct 2023. To prevent impact till the fix is rolled out we are applying short term mitigation to all caches which will reduce the log file size. The next update will be provided by 19:00 UTC on 8 Oct 2023 or as events warrant, to allow time for the short term mitigation to progress.
Impact statement: Beginning as early as 11 Aug 2023, you have been identified as a customer experiencing timeouts and high server load for smaller size caches (C0/C1/C2).
Current status: Investigation revealed the cause to be a change in behavior of one of the Azure security monitoring services agent used by Azure Cache for Redis. Monitoring Agent subscribes to the event log and has scheduled backoff for resetting subscription in case no events are received. In some cases scheduled backoff is not working as expected and can increase the frequency of subscription resetting which can significantly affect CPU usage for smaller size caches. Currently, we are in progress of rolling out of the hotfix to the impacted regions which is 80% completed. Initially we estimated this to complete by 11 Oct 2023, however, progress shows we are expected to complete by 09 Oct 2023. To prevent impact till the fix is rolled out we are applying short term mitigation to all caches which will reduce the log file size. The next update will be provided by 19:00 UTC on 8 Oct 2023 or as events warrant, to allow time for the short term mitigation to progress.
Summary of Impact: Between as early as 11 Aug 2023 and 18:00 UTC on 8 Oct 2023, you were identified as a customer who may have experienced timeouts and high server load for smaller size caches (C0/C1/C2).
Current Status: This issue is now mitigated. More information will be provided shortly.
What happened?
Between as early as 11 Aug 2023 and 18:00 UTC on 8 Oct 2023, you were identified as a customer who may have experienced timeouts and high server load for smaller size caches (C0/C1/C2).
What do we know so far?
We identified a change in behavior of one of the Azure security monitoring services agents used by Azure Cache for Redis. Monitoring Agent subscribes to the event log and has scheduled backoff for resetting subscription in case no events are received. In some cases, scheduled backoff is not working as expected and can increase the frequency of subscription resetting which can significantly affect CPU usage for smaller size caches.
How did we respond?
To address this issue, engineers performed manual action on the underlying Virtual Machines of impacted caches. After further monitoring, internal telemetry confirmed this issue is mitigated and full-service functionality was restored.
What happens next?
We will continue to investigate to establish the full root cause and prevent future occurrences. Stay informed about Azure service issues by creating custom service health alerts: https://aka.ms/ash-videos for video tutorials and https://aka.ms/ash-alerts for how-to documentation.
Summary of Impact: Between 20:30 UTC on 18 Aug. 23 and 05:10 UTC on 19 Aug. 23, you were identified as customer using Workspace-based Application Insights resources who may have experienced 7-10% data gaps during the impact window, and potentially incorrect alert activations.
Preliminary Root Cause: We identified that the issue was caused due to a code bug as part of the latest deployment which has caused some drop in the data.
Mitigation: We have rolled back the deployment to last known good build to mitigate the issue.
Additional Information: Following additional recovery efforts, we have re-ingested the data that was not correctly ingested due to this event, after further investigation it was discovered that the initial re-ingested data had incorrect TimeGenerated values, instead of the original TimeGenerated value. This may cause incorrect query results which may further cause incorrect alerts or report generation. We have investigated the issue that caused this behavior so future events utilizing data recovery processes will re-ingest the data with the correct, original TimeGenerated value.
If you need any further assistance on this, please raise the support ticket for the same.
Summary of Impact: Between 20:30 UTC on 18 Aug. 23 and 05:10 UTC on 19 Aug. 23, you were identified as customer using Workspace-based Application Insights resources who may have experienced 7-10% data gaps during the impact window, and potentially incorrect alert activations.
Preliminary Root Cause: We identified that the issue was caused due to a code bug as part of the latest deployment which has caused some drop in the data.
Mitigation: We have rolled back the deployment to last known good build to mitigate the issue.
Additional Information: Following additional recovery efforts, we re-ingested the data that was not correctly ingested due to this event, after further investigation it was discovered that the initial re-ingested data had incorrect TimeGenerated values, instead of the original TimeGenerated value. This may have caused incorrect query results which may have further caused incorrect alerts or report generation. Our investigation extended past previous mitigation and we were able to identify a secondary code bug that caused this behavior. We deployed a hotfix using our Safe Deployment Procedures that re-ingested the data with the correct, original TimeGenerated value. All regions are now recovered and previously incorrect TimeGenerated values are now corrected.
If you need any further assistance on this, please raise the support ticket for the same.
Next Steps: We will continue to investigate to establish the full root cause and prevent future occurrence. Stay informed about Azure service issues by creating custom service health alerts: https://aka.ms/ash-videos for video tutorials and https://aka.ms/ash-alerts for how-to documentation.
