Free Trial

Safari Books Online is a digital library providing on-demand subscription access to thousands of learning resources.

  • Create BookmarkCreate Bookmark
  • Create Note or TagCreate Note or Tag
  • PrintPrint

Checklist

Table 5-1 describes some potential problem areas with file systems, with suggestions on how you can use DTrace to troubleshoot them.

Table 5-1. File System I/O Checklist
IssueDescription
VolumeApplications may be performing a high volume of file system I/O, which could be avoided or optimized by changing their behavior, for example, by tuning I/O sizes and file locations (tmpfs instead of nfs, for example). The file system may break up I/O into multiple physical I/O of smaller sizes, inflating the IOPS. DTrace can be used to examine file system I/O by process, filename, I/O size, and application stack trace, to identify what files are being used, how, and why.
LatencyA variety of latencies should be examined when analyzing file system I/O:
  • Disk I/O wait, for reads and synchronous writes

  • Locking in the file system

  • Latency of the open() syscall

  • Large file deletion time

Each of these can be examined using DTrace.
QueueingUse DTrace to examine the size and wait time for file system queues, such as queueing writes for later flushing to disk. Some file systems such as ZFS use a pipeline for all I/O, with certain stages serviced by multiple threads. High latency can occur if a pipeline stage becomes a bottleneck, for example, if compression is performed; this can be analyzed using DTrace.
CachesFile system performance can depend on cache performance: File systems may use multiple caches for different data types (directory names, inodes, metadata, data) and different algorithms for cache replacement and size. DTrace can be used to examine not just the hit and miss rate of caches, but what types of data are experiencing misses, what contents are being evicted, and other internals of cache behavior.
ErrorsThe file system interface can return errors in many situations: invalid file offsets, permission denied, file not found, and so on. Applications are supposed to catch and deal with these errors with them appropriately, but sometimes they silently fail. Errors returned by file systems can be identified and summarized using DTrace.
ConfigurationFile access can be tuned by flags, such as those on the open() syscall. DTrace can be used to check that the optimum flags are being used by the application, or if it needs to be configured differently.



  

You are currently reading a PREVIEW of this book.

                                                                                                                    

Get instant access to over $1 million worth of books and videos.

  

Start a Free Trial


  
  • Safari Books Online
  • Create BookmarkCreate Bookmark
  • Create Note or TagCreate Note or Tag
  • PrintPrint