Loading...
Loading...
CPU utilization remains consistently high, often exceeding 80%, even during periods of low throughput and minimal active streams. Elevated system (%system) CPU usage observed in SAR statistics. High system load averages that do not correlate with backup or replication workload. DDFS processes consume substantially more CPU resources than expected. Performance analysis shows significant activity in READDIRPLUS-related code paths. Large numbers of OSTREADDIRPLUS or ddp_readdir operations observed in Autosupport data. Backup, restore, or file access operations may become slower than expected. Increased active TCP connection counts and possible connection rejection alerts. Frequent "5006-end of directory" messages may be observed in BoostFS or DDFS logs. One dd_guts scheduler thread may appear heavily utilized while other scheduler threads remain comparatively idle. NFS RPC layer statistics may show elevated data_wait and svc_func values.
Analysis indicated that excessive READDIR and READDIRPLUS operations originating from BoostFS-mounted file systems were contributing to elevated CPU utilization. Historical investigations have shown that automated indexing, security scanning, monitoring utilities, and file system traversal processes can generate large volumes of directory enumeration requests, resulting in increased DDFS and kernel CPU usage even when backup workloads remain relatively low. Common sources of excessive READDIR activity include: mlocate/updatedb indexing processes scanning BoostFS mount points. Security and antivirus scanning applications. Monitoring and health-check tools polling mounted file systems. Frequent file system enumeration activities such as recursive directory listings and stat operations. Applications using BoostFS or SDK interfaces that repeatedly traverse large directory structures. Automated scripts performing continuous file system discovery or inventory collection. These activities can generate millions of READDIR operations per day, creating significant processing overhead within DDFS and the underlying file system.
Review systems with BoostFS mounts for applications that perform frequent directory enumeration, indexing, monitoring, or scanning activities. Recommended actions include: Identify applications generating excessive READDIR activity against BoostFS-mounted file systems. Exclude BoostFS mount points from mlocate/updatedb indexing by updating the PRUNEFS configuration. Review and limit unnecessary antivirus scans, security scans, monitoring tools, health-check processes, and file system polling activities targeting BoostFS mounts. Reduce or eliminate frequent recursive directory traversal operations where possible. Monitor OSTREADDIRPLUS , ddp_readdir , CPU utilization, and load averages after implementing changes. Follow Data Domain and BoostFS best practices to prevent scanner and indexing applications from continuously traversing mounted backup file systems. Historical investigations have shown a substantial reduction in CPU utilization after BoostFS mount points were excluded from automated cron jobs performing indexing and scanning activities. This behavior indicates that excessive READDIR activity generated by these processes can be a significant contributor to elevated DDFS CPU utilization. Additionally, frequent "5006-end of directory" messages may indicate repeated directory read operations. The volume of these messages can be used as an indicator of the amount of READDIR activity occurring against the Data Domain system. Read operations on BoostFS-mounted file systems are typically more expensive than write operations, making it important to avoid unnecessary scanning or traversal of mounted backup namespaces. Check cron jobs on boostfs clients: ls -l /etc/cron.daily
Click on a version to see all relevant bugs
Dell Integration
Learn more about where this data comes from
BugZero Plan
Streamline upgrades with automated vendor bug scrubs
BugZero Prevent
Wish you caught this bug sooner? Get proactive today.