-
Notifications
You must be signed in to change notification settings - Fork 4.1k
cloud/s3: audit memory usage during uploads #115196
Copy link
Copy link
Open
Labels
A-disaster-recoveryC-investigationFurther steps needed to qualify. C-label will change.Further steps needed to qualify. C-label will change.O-supportWould prevent or help troubleshoot a customer escalation - bugs, missing observability/tooling, docsWould prevent or help troubleshoot a customer escalation - bugs, missing observability/tooling, docsP-3Issues/test failures with no fix SLAIssues/test failures with no fix SLAT-disaster-recovery
Description
Metadata
Metadata
Assignees
Labels
A-disaster-recoveryC-investigationFurther steps needed to qualify. C-label will change.Further steps needed to qualify. C-label will change.O-supportWould prevent or help troubleshoot a customer escalation - bugs, missing observability/tooling, docsWould prevent or help troubleshoot a customer escalation - bugs, missing observability/tooling, docsP-3Issues/test failures with no fix SLAIssues/test failures with no fix SLAT-disaster-recovery
In #115190 we observed 600+ mb of heap allocated memory due to the s3 sdk buffers. With a concurrency of 5 and chunk size of 8mb, this number did not initially make sense to us. We should dig in more here to see if we're configuring something wrong.
Additional context in thread leading up to https://cockroachlabs.slack.com/archives/C05P9KPRZFW/p1701175750732169?thread_ts=1701123237.549029&cid=C05P9KPRZFW
Jira issue: CRDB-33929