Poor GFS2 write performance for large single files on Red Hat Enterprise Linux 5

Solution Verified - Updated -

Issue

  • Copying a big file to GFS2 file system needs more time than ext3, and these device are backed by the same SAN disk.
  • We were trying to create a 400 Gb file in a GFS2 share. It was running at around 50Mb/s approx. Suddenly, after copying 80%, the rate dropped to around 30Kb/s and stayed like that. We tried to kill the process but couldn't (which is normal) and after a few minutes it was killed. Then tried again after a few minutes and it was successfully created at approximately 50Mb/s.
  • The performance on GFS2 volumes decrease constantly and we have to unmount and remount them every day to keep our application running.
  • GFS2 copying to backup hangs or drops to very slow write speeds after the kernel was updated to 2.6.18-238.1.1.el5.
  • First we create the backup files locally, 7 files 16Gb and 1 file 3Gb, the last step is copy the local files from ext3 (not from GFS2 filesystem) dir to the GFS2.
  • GFS2 with NFS becomes very sluggish when transferring large files to the SAN via NFS v3 or v4 (files larger than 20Gb) the whole system becomes very sluggish.
  • GFS2 write performance degrades after ~500Gb copy.

Environment

  • Red Hat Enterprise Linux 5
  • Kernels prior to 2.6.18-274.el5 (most common on 2.6.18-238*.el5 versions, but possible on all earlier versions)
  • Red Hat Cluster Suite
  • Red Hat Resilient Storage
  • Global File System 2 (GFS2)
  • Writing a large file (usually several gigabytes) to the GFS2 filesystem

Subscriber exclusive content

A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.

Current Customers and Partners

Log in for full access

Log In

New to Red Hat?

Learn more about Red Hat subscriptions

Using a Red Hat product through a public cloud?

How to access this content