GFS2 much slower than ext3 for creating Oracle forms and reports on Red Hat Enterprise Linux 5
Issue
- I have a 4-node cluster (physical server) with a GFS file system. When running jobs to compile Oracle forms and reports, it is terribly slow when compared to our test server (single VM machine).
- I then proceed to test the performance of the GFS cluster by removing 3-nodes from the cluster and mount the single node GFS filesystems with lock_nolock in addition to the existing option - noatime and nodiratime.
- Mounting with lock_nolock proves to be a lot faster as compared to running with all 4-nodes.
- What can be done here to improve the performance of the GFS filesystem?
Environment
- Red Hat Enterprise Linux Server 5 (with Resilient Storage Add Ons)
- Red Hat High Availability cluster with 2 or more nodes
- GFS2 filesystem mounted on all cluster nodes
- GFS2 filesystems are mounted with
defaults,noatime,nodiratime,acl,quota=off
- There is no entry for plock_rate_limit in /etc/cluster/cluster.conf, so the default of 100 plocks-per-second is used.
- GFS2 filesystems are mounted with
- Workload consists of Oracle Application AD Administration (Version 12.0.0)
- This application makes extensive use of Posix locks (plocks).
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.