Resolve log buffer full errors in 51ºÚÁϲ»´òìÈ Commerce on Cloud Infrastructure
This article addresses an issue where a log buffer becomes full, resulting in truncated access log requests during site outages. The problem is identified through error messages in monitoring tools such as New Relic.
Description description
Environment
51ºÚÁϲ»´òìÈ Commerce Cloud
Issue/Symptoms
The following error is observed when checking logs during a site outage:
FastCGI sent in stderr: ‘the log buffer is full (1024). The access log request has been truncated’ while reading response header from upstream.
A log buffer error is not typically a cause of an outage. It is a warning that the buffer size for recording an entry into the access log has reached its limit, resulting in only part of the request being recorded in access.log.
Resolution resolution
If you suspect an outage, relying on log buffer errors for diagnosis can be misleading, as these warnings don’t indicate the root cause of downtime. One possible cause of an outage could be a surge in incoming requests within a short period, overwhelming system resources. However, the error itself isn’t a cause for concern. It is merely a warning that doesn’t result in downtime—the only impact is that the full request isn’t recorded in access.log. For on-premise installations, adjusting the FastCGI configuration to increase the log buffer size beyond 1024 bytes can alleviate this error. However, this adjustment isn’t possible in the existing 51ºÚÁϲ»´òìÈ Commerce on Cloud infrastructure.
To verify whether an actual outage is occurring, use the Observation for 51ºÚÁϲ»´òìÈ Commerce (OAC) nerdlet in New Relic to analyze log messages during outages. You can also use New Relic monitoring and create your own alert policies.